def overlap_contours(i,j): # Take the 'j'th contour, from the contours_sorted array, and the for every point of the 'j' check if its inside 'i' or on 'i' or outside 'i' # If we find that the contour 'j' is conpletely outside, we return -1, else if partially overlapped, we return 0, and if completely subsumed, we return +1 completely_outside = True completely_subsumed = True for k in contours[j][0]: #for every point in contour 'j' k_loc = cv2.pointPolygonTest(contours[i],(k[0],k[1]),False) if k_loc == 0: #Its on the contour[i] completely_outside = False completely_subsumed = False break elif k_loc == -1: #Its not subsumed completely_subsumed = False # end of for if completely_subsumed == True: return(j) if completely_outside == False and completely_subsumed == False: return(-1) #on the border if completely_outside == True: # we test the reverse condition, thats the object 'i' might be subsumed completely within object 'i' completely_subsumed = True for k in contours[i][0]: # for every point in contour 'i' if cv2.pointPolygonTest(contours[j], (k[0], k[1]), False) < 1: # Its on or outside the contour[j] completely_subsumed = False break if completely_subsumed == True: return(i) else: return(-2) return(-1)
def distance_image(self,point=None):
'''find the distance between a point and adjacent point on contour specified. Point should be a tuple or list (x,y)
If no point is given, distance to all point is calculated and distance image is returned'''
if type(point) == tuple:
if len(point)==2:
self.dist = cv2.pointPolygonTest(self.cnt,point,True)
return self.dist
else:
dst = np.empty(self.img.shape)
for i in xrange(self.img.shape[0]):
for j in xrange(self.img.shape[1]):
dst.itemset(i,j,cv2.pointPolygonTest(self.cnt,(j,i),True))
dst = dst+127
dst = np.uint8(np.clip(dst,0,255))
# plotting using palette method in numpy
palette = []
for i in xrange(256):
if i<127:
palette.append([2*i,0,0])
elif i==127:
palette.append([255,255,255])
elif i>127:
l = i-128
palette.append([0,0,255-2*l])
palette = np.array(palette,np.uint8)
self.h2 = palette[dst]
return self.h2
def isEllipse(contour):
'''
Detects if the given polygon is an ellipse.
Returns the ellipse form of the polygon if it's an ellipse, None otherwise.
'''
try: rect = cv2.fitEllipse(contour) # Rotated rectangle representing the ellipse it tries to fit
except: return None
(x, y), (w, h), angle = rect # x offset, y offset, width, height, angle
w, h = h, w # Switch them since our ellipses are usually rotated 90 degrees
TEST_POINTS = 16
E = 0.1 # Radius percentage increase/decrease for test points
MIN_SUCC_RATE = 0.9
successes = 0
for a in np.arange(0, 2*math.pi, 2*math.pi/TEST_POINTS):
ipoint = (x+0.5*w*math.cos(a)*(1-E), y+0.5*h*math.sin(a)*(1-E))
opoint = (x+0.5*w*math.cos(a)*(1+E), y+0.5*h*math.sin(a)*(1+E))
test = cv2.pointPolygonTest(contour, ipoint, False)
if cv2.pointPolygonTest(contour, ipoint, False) > 0: # The inside point is inside
successes += 1
if cv2.pointPolygonTest(contour, opoint, False) < 0: # The outside point is outside
successes += 1
succ_rate = successes / (TEST_POINTS * 2)
if succ_rate >= MIN_SUCC_RATE:
return rect
else:
return None
def ppt(ct, p, b): if type(p) == tuple: return -cv2.pointPolygonTest(ct, p, b) elif len(np.shape(p)) == 1: return -cv2.pointPolygonTest(ct, tuple(p), b) else: return np.array([-cv2.pointPolygonTest(ct, tuple(pp), b) for pp in np.array(p)])
def plot_intercontour_hist(image, outer_contour_id, contours, graph, normalized=True):
outer_contour = contours[outer_contour_id]
(x, y, width, height) = cv2.boundingRect(outer_contour)
subimage = get_subimage(image, (x, y), (x + width, y + height))
monochrome = cv2.cvtColor(subimage, cv2.COLOR_BGR2GRAY)
inverted_mask = cv2.compare(monochrome, monochrome, cv2.CMP_EQ)
inner_contours = [contours[int(contour_id)] for contour_id in graph.successors(outer_contour_id)]
for i in range(width):
for j in range(height):
point = (x + i, y + j)
outer_contour_test = cv2.pointPolygonTest(outer_contour, point, 0)
inner_contours_test = -1
for inner_contour in inner_contours:
inner_contour_test = cv2.pointPolygonTest(inner_contour, point, 0)
if inner_contour_test > 0:
inner_contours_test = 1
break
if outer_contour_test >= 0 and inner_contours_test < 0:
inverted_mask[j][i] = 0
mask = cv2.bitwise_not(inverted_mask)
cv.Set(cv.fromarray(subimage), WHITE, cv.fromarray(inverted_mask))
inner_contour_id = len(str(inner_contours))
print 'inner contour id: ', inner_contour_id
image_name = '%d-%s'%(outer_contour_id, inner_contour_id)
#cv2.imshow(image_name, subimage)
#subhists = plot_hist(subimage, mask, image_name)
(subhists, winnames) = plot_hist_hls(subimage, mask, image_name, normalized)
return subhists, subimage, mask, x, y, winnames
def isEllipse(contour):
rect = cv2.fitEllipse(contour) # Rotated rectangle representing the ellipse it tries to fit
(x, y), (w, h), angle = rect # x offset, y offset, width, height, angle
w, h = h, w # Switch them since our ellipses are usually rotated 90 degrees
# Draw TEST_INSIDE points inside the ellipse and TEST_OUTSIDE points outside and see if they're in the hull
TEST_INSIDE = 10
TEST_OUTSIDE = 10
# Equation of ellipse: (x/a)^2 + (y/b)^2 = 1
e = 0.1
tests = []
tests.append(cv2.pointPolygonTest(contour, (x, y), False))
tests.append(cv2.pointPolygonTest(contour, (x+w*(0.5-e), y), False))
tests.append(cv2.pointPolygonTest(contour, (x-w*(0.5-e), y), False))
tests.append(cv2.pointPolygonTest(contour, (x, y+h*(0.5-e)), False))
tests.append(cv2.pointPolygonTest(contour, (x, y-h*(0.5-e)), False))
tests.append(not cv2.pointPolygonTest(contour, (x+w*(0.5-e), y+h*(0.5-e)), False))
tests.append(not cv2.pointPolygonTest(contour, (x+w*(0.5-e), y-h*(0.5-e)), False))
tests.append(not cv2.pointPolygonTest(contour, (x-w*(0.5-e), y+h*(0.5-e)), False))
tests.append(not cv2.pointPolygonTest(contour, (x-w*(0.5-e), y-h*(0.5-e)), False))
for test in tests:
if test == -1.0:
return False
return True
def shrinkBoundingRectToCircumscribed(maskHull, x, y, w, h):
# now we are going to SHRINK it until all 4 corners fit
# ul= np.array([0,0])
# ur = np.array([0,0])
# ll = np.array([0,0])
# lr = np.array([0,0])
# ul= np.zeros(2,dtype = np.int32)
# ur = np.zeros(2,dtype = np.int32)
# ll = np.zeros(2,dtype = np.int32)
# lr = np.zeros(2,dtype = np.int32)
if (maskHull is None):
return (0,0,0,0)
testPolygon = maskHull
# print "test polygon is:"
# print testPolygon
while True:
# ul[0]=x
# ul[1]=y
# ur[0]=x+w
# ur[1]=y
# ll[0]=x
# ll[1]=y+h
# lr[0]=x+2
# lr[1]=y+h
ul = (x, y)
ur = (x + w - 1, y)
ll = (x, y + h - 1)
lr = (x + w - 1, y + h - 1)
# ul = (y,x)
# ur = (y,x+w)
# ll = (y+h,x)
# lr = (y+h,x+w)
testval = (cv2.pointPolygonTest(testPolygon, ul, False) == 1) and (
cv2.pointPolygonTest(testPolygon, ur, False) == 1) and (cv2.pointPolygonTest(testPolygon, ll, False) == 1) and (
cv2.pointPolygonTest(testPolygon, lr, False) == 1)
if (testval):
# print "matches at:"
# print ul, ur, ll, lr
return (x, y, w, h)
# print "Testing %d,%d size %d,%d", (x,y,x+w,y+h)
x = x + 1
y = y + 1
w = w - 2
h = h - 2
if (w <= 0 or h <= 0):
return (0, 0, 0, 0)
pass
def get_targets(self, fly_erased_img, headpoint, centroidpoint, _bodyAxis):
kernel = np.ones((5,5),np.uint8)
_, mask = cv2.threshold(fly_erased_img, 60, 255, cv2.THRESH_BINARY)
mask = cv2.erode(mask, kernel, iterations=1)
contourImage = mask.copy()
contourImage = np.pad(contourImage,((2,2),(2,2)), mode='maximum')
contours, hierarchy1 = cv2.findContours(contourImage, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
hierarchy = hierarchy1[0]
for x in hierarchy:
if x[3] <0:
parent = x
#headpoint = (int(track.loc[framenumber, 'c_head_location_x']), int(track.loc[framenumber, 'c_head_location_y']))
candidateTargets = []
for component in zip(contours, hierarchy):
c = component[0]
h = component[1]
centroidCheck = cv2.pointPolygonTest(c,centroidpoint,True)
if centroidCheck <=0:
if np.array_equal(hierarchy[h[3]], parent) : #is in outer hierarchy (parent is edge.)
if h[2] > 0: # has child (targets have inner and outer edge)
if (cv2.contourArea(c) <= 150000) & (cv2.contourArea(c) >= 20000):
ellipse = cv2.fitEllipse(c)
if not self.pointInEllipse(centroidpoint[0],centroidpoint[1],ellipse[0][0],ellipse[0][1],ellipse[1][0],ellipse[1][1],ellipse[2]):
candidateTargets.append(c)
areas = []
if len(candidateTargets) >0:
for T in range(len(candidateTargets)):
areas.append(cv2.contourArea(candidateTargets[T]))
TARGET = cv2.convexHull(candidateTargets[areas.index(max(areas))] )
M = cv2.moments(TARGET)
targCentre = (int(M['m10']/M['m00']), int(M['m01']/M['m00']))
distance = -1.0*cv2.pointPolygonTest(TARGET,headpoint,True) / 135.5 # based on 135.5 pixels per mm
angle= self.angle_from_vertical(headpoint, targCentre)
approachAngle= angle - _bodyAxis #track.loc[framenumber, 'd_bodyAxis']
if approachAngle < 0:
approachAngle *= -1.0
if approachAngle >=180.0:
approachAngle -= 180.0
else:
distance = np.nan
approachAngle = np.nan
TARGET = None
return TARGET, distance, approachAngle
def triangle_is_valid(self, start):
# check if middle point of triangle is inside polygon. otherwise triangle is invalid
middle = self.calc_triangle_middle(self.triangle)
if cv2.pointPolygonTest(np.array([self.points], 'int32'), middle, False) < 0:
return False
# check if other points pf polygon are not inside triangle. otherwise triangle is invalid
for i, point in enumerate(self.points):
# ignore points of triangle
if i < start or i > start + 1:
if cv2.pointPolygonTest(np.array([self.triangle], 'int32'), (point[0], point[1]), False) > 0:
return False
# check if area of triangle is not too big and not too small for polygon
area = cv2.contourArea(np.array(self.triangle, 'int32'))
return MIN_COLOR_TRIANGLE_AREA < area / self.area < MAX_COLOR_TRIANGLE_AREA
def __compare_segmentations(img_a, img_b):
# image, contours, hierarchy = cv.findContours(image, mode, method[, contours[, hierarchy[, offset]]]) [opencv3]
# contours, hierarchy = cv.findContours(image, mode, method[, contours[, hierarchy[, offset]]]) [opencv2]
out_img_a, ct_img_a, h_a = cv.findContours(img_a, cv.RETR_TREE, cv.CHAIN_APPROX_NONE)
out_img_b, ct_img_b, h_b = cv.findContours(img_b, cv.RETR_TREE, cv.CHAIN_APPROX_NONE)
error = []
height, width = img_a.shape[:2]
for comp_a in ct_img_a:
mask_a = np.zeros((height, width), np.uint8)
mask_b = np.zeros((height, width), np.uint8)
first_point = comp_a[0][0]
cv.fillConvexPoly(mask_a, comp_a, (255))
for comp_b in ct_img_b:
if cv.pointPolygonTest(comp_b, (first_point[0], first_point[1]), False) >= 0:
cv.fillConvexPoly(mask_b, comp_b, (50))
break
set_difference = mask_a - mask_b
set_difference_count = (set_difference == 255).sum()
error.append(set_difference_count)
return error
def rotate_contour(contour, frame_bin):
center, size, angle = cv2.minAreaRect(contour)
xt,yt,h,w = cv2.boundingRect(contour)
region_points = []
for i in range (xt,xt+h):
for j in range(yt,yt+w):
dist = cv2.pointPolygonTest(contour,(i,j),False)
if dist>=0 and frame_bin[j,i]==255: # da li se tacka nalazi unutar konture?
region_points.append([i,j])
cx,cy = center
height, width = size
if width<height:
angle-=90
# Rotiranje svake tačke regiona oko centra rotacije
alpha = np.pi/2 - abs(np.radians(angle))
region_points_rotated = np.ndarray((len(region_points), 1,2), dtype=np.int32)
for i, point in enumerate(region_points):
x = point[0]
y = point[1]
#TODO 1 - izračunati koordinate tačke nakon rotacije
rx = np.sin(alpha)*(x-cx) - np.cos(alpha)*(y-cy) + cx
ry = np.cos(alpha)*(x-cx) + np.sin(alpha)*(y-cy) + cy
region_points_rotated[i] = [rx,ry]
return region_points_rotated
def select_and_convexify(self, selection_method, make_convex, object):
# index_desired = self.candidate_objects.index(self.desired_object)
object_index = self.candidate_objects.index(object)
self.segments[object] = np.zeros(self.segmentation.shape, np.uint8)
self.contours, hierarchy = cv2.findContours(np.logical_and(self.bin_mask, self.segmentation == object_index).astype('uint8'), cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
if self.contours:
# http://opencv-python-tutroals.readthedocs.org/en/latest/py_tutorials/py_imgproc/py_contours/py_contour_features/py_contour_features.html
# compute the area for each contour
if selection_method == "all":
self.selected_contours = range(len(self.contours))
elif selection_method == "largest":
contour_areas = [cv2.contourArea(cnt) for cnt in self.contours]
self.selected_contours = [np.argmax(contour_areas)]
elif selection_method == "max_smooth":
# find most likely point for object on the parts that were segmented
argmax = (self.posterior_images_smooth[object] * (self.segmentation == object_index)).argmax()
# compute the image coordinates of this point
max_posterior_point = np.unravel_index(argmax, self.posterior_images_smooth[object].shape)[::-1]
# select contour that includes this point
self.selected_contours = [np.argmax([cv2.pointPolygonTest(cnt,max_posterior_point,False) for cnt in self.contours])]
else:
print("I DON'T KNOW THIS SELECTION METHOD!")
for i in self.selected_contours:
if make_convex:
self.contours[i] = cv2.convexHull(self.contours[i])
cv2.drawContours(self.segments[object], [self.contours[i]], 0, 255, -1)
def getInternalPixels(img, contour, nullval=-1):
maxx, maxy = 0, 0
minx, miny = sys.maxsize, sys.maxsize
for val in contour:
if val[0][0] > maxx :
maxx = val[0][0]
if val[0][0] < minx :
minx = val[0][0]
if val[0][1] > maxy :
maxy = val[0][1]
if val[0][1] < miny :
miny = val[0][1]
retPx = np.zeros((maxy-miny+3, maxx-minx+3, 3), np.uint8)
ri, ci = 0, 0
for row in range(miny-1, maxy+2):
ci = 0
for col in range(minx-1, maxx+2):
if cv2.pointPolygonTest(contour, (col, row), False) >= 0:
retPx[ri][ci] = img[row][col]
else:
retPx[ri][ci] = nullval
ci += 1
ri += 1
return retPx
def extract_face(face_id, img_width, img_height, face_name):
landmark = api.detection.landmark(face_id=face_id, type='83p')
#center point to expend the contour
nose_x = landmark['result'][0]['landmark']['nose_tip']['x']
nose_y = landmark['result'][0]['landmark']['nose_tip']['y']
nose_x = nose_x / 100 * img_width
nose_y = nose_y / 100 * img_height
contour = []
contour_point_name = []
for v in contour_name_list2:
x = landmark['result'][0]['landmark'][v]['x']
y = landmark['result'][0]['landmark'][v]['y']
x = x / 100 * img_width
y = y / 100 * img_height
if 'mouth' in v:
dx = (x-nose_x) * 1.2
dy = (y-nose_y) * 1.2
#print '%s: %5.2f %5.2f' % (v, x-nose_x, y-nose_y)
x = dx + nose_x
y = dy + nose_y
contour.append([x, y])
contour = np.array(contour, dtype=np.int32)
extract_face = cv2.imread(face_name)
#print extract_face.shape
for x in xrange(img_width):
for y in xrange(img_height):
if cv2.pointPolygonTest(contour,(x,y),False) < 0:
#print x, y
extract_face[y][x][0] = 0
extract_face[y][x][1] = 0
extract_face[y][x][2] = 0
return extract_face
def pointInRoi(self, point):
"""
Returns True if the point is in the roi
:param tuple point: and (x,y) point
"""
return cv2.pointPolygonTest(self.points, point, False) > 0
def testInside(self, p, measureDist=False):
"""extends Widget.testInside() to support distance to ellipse.
Note: distance to ellipse is approximate."""
if self.type == 'ellipse':
pdiff = -(self.center-p)
uc = numpy.dot(pdiff, nu) / self.u
vc = numpy.dot(pdiff, nv) / self.v
d2 = uc*uc + vc*vc
if measureDist:
d = math.sqrt(d2)
if d:
uc /= d
vc /= d
else:
uc = 1
vc = 0
pc = self.center + uc*self.u*self.nu + vc*self.v*self.nv
diff = -(pc - p)
dc = math.sqrt(numpy.dot(diff, diff))
if d2 < 1:
return dc
else:
return -dc
elif d2 == 1:
return 0
elif d2 < 1:
return 1
elif d2 > 1:
return -1
else:
if len(self.points) != 9:
updatePoints()
ppoly = self.points[self.outer]
return cv2.pointPolygonTest( ppoly, a2t(p), measureDist )
def find_morse_contours(group_morse):
h, w = group_morse.shape[:2]
group_morse = cv2.erode(group_morse, np.ones((11, 11), np.uint8), iterations=8)
# erode to create big contours - groups, aka letters
group_morse[0:0.03 * h, :] = 255
group_morse[0.97 * h:h, :] = 255
group_morse[:, 0:0.03 * w] = 255
group_morse[:, 0.97 * w:w] = 255
_, grp_cnts, _ = cv2.findContours(group_morse.copy(), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
morse_grp_cnts = []
for c in grp_cnts:
if cv2.arcLength(c, 1) < ((2 * h + 2 * w) * 0.5):
morse_grp_cnts.append(c)
morse_groups = []
morse_groups_cent = []
for grp_cnt in morse_grp_cnts:
grp = []
grp_cent = []
for i in range(0, len(morse_cnts)):
if cv2.pointPolygonTest(grp_cnt, (morse_cent[i][0], morse_cent[i][1]), False) > 0:
# if centroid is in group shape, this dot/dash belongs to group
grp.append(morse_cnts[i])
grp_cent.append(morse_cent[i])
if len(grp) != 0:
morse_groups.append(grp) # add group to group set
morse_groups_cent.append(grp_cent)
morse_groups = morse_groups[::-1]
morse_groups_cent = morse_groups_cent[::-1]
for i in range(len(morse_groups)):
morse_groups[i], morse_groups_cent[i] = sort_by_centroids(morse_groups[i], morse_groups_cent[i])
return group_morse, morse_groups, morse_groups_cent, morse_grp_cnts
def identifyNotes(img,noteheads,staffData): # Invert the image for findContours img = 255 - img # Find contours contours,hierarchy = cv2.findContours(img,cv2.RETR_TREE,cv2.CHAIN_APPROX_NONE) # Remove contours that have dimensions that are too large to correspond to notes heightThreshold = 8*(staffData.lineThickness + staffData.lineSpacing) filteredContours = [] for contour in contours: roi = cv2.boundingRect(contour) if (roi[3] < heightThreshold): filteredContours.append(contour) contours = filteredContours # contourMap is a list of tuples (contour,noteheads) where contour is a list of points representing a contour and noteheads is a list of noteheads whose centre lies within that contour contourMap = [] # Identify, for each contour, which noteheads has a centre that lies within that contour for contour in contours: noteheadsInContour = [] for notehead in noteheads: if (cv2.pointPolygonTest(contour,tuple(map(lambda (x,y): x+y/2,zip(notehead.point,notehead.dimensions))),False) == 1): noteheadsInContour.append(notehead) if (noteheadsInContour): contourMap.append((contour,noteheadsInContour))
def findMaxContours():
conts = []
for c in range(180):
# for c in [170,50,86,111]:
lower_blue = np.array([c-hdelta,0,0])
upper_blue = np.array([c+hdelta,255,250])
mask = cv2.inRange(img, lower_blue, upper_blue)
x = kersize
kernel = np.ones((x,x), np.uint8)
erosion = cv2.dilate(mask,kernel,iterations = 1)
kernel = np.ones((x/2,x/2), np.uint8)
erosion = cv2.erode(erosion,kernel,iterations = 1)
contours,h = cv2.findContours(erosion,1,2)
for i,cnt in enumerate(contours):
x,y,w,h = cv2.boundingRect(cnt)
if w > 0.6 * img.shape[1]:
app = True
for cnt2 in conts:
if cv2.pointPolygonTest(cnt2, tuple(cnt2[0][0]) , False):
app = False
if app:
conts.append(cnt)
contsS = sorted(conts, key=lambda x: cv2.boundingRect(x)[2] )
for i,cnt in enumerate(contsS[-90:] ):
cv2.drawContours(img,[cnt],0,(0,i*10 % 255, (i*223)%255 ),4)
cv2.imshow('dst',img)
return contsS[-100:]
def select_roi2(image_orig, image_bin):
img, contours_borders, hierarchy = cv2.findContours(image_bin.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)
contours = []
contour_angles = []
contour_centers = []
contour_sizes = []
for contour in contours_borders:
if((cv2.contourArea(contour) >200) and (isTriangleP(contour) or isOsmougao(contour) or isSquareP(contour) or isCircleP(contour))):
center, size, angle = cv2.minAreaRect(contour)
xt,yt,h,w = cv2.boundingRect(contour)
cv2.rectangle(image_orig,(xt,yt),(xt+w,yt+h),(255,0,0),10)
region_points = []
for i in range (xt,xt+h):
for j in range(yt,yt+w):
dist = cv2.pointPolygonTest(contour,(i,j),False)
if dist>=0 and image_bin[j,i]==255: # da li se tacka nalazi unutar konture?
region_points.append([i,j])
contour_centers.append(center)
contour_angles.append(angle)
contour_sizes.append(size)
contours.append(region_points)
#Postavljanje kontura u vertikalan polozaj
contours = rotate_regions(contours, contour_angles, contour_centers, contour_sizes)
#spajanje kukica i kvacica
#contours = merge_regions(contours)
regions_dict = {}
for contour in contours:
min_x = min(contour[:,0])
max_x = max(contour[:,0])
min_y = min(contour[:,1])
max_y = max(contour[:,1])
region = np.zeros((max_y-min_y+1,max_x-min_x+1), dtype=np.int16)
for point in contour:
x = point[0]
y = point[1]
# TODO 3 - koordinate tacaka regiona prebaciti u relativne koordinate
'''Pretpostavimo da gornja leva tačka regiona ima apsolutne koordinate (100,100).
Ako uzmemo tačku sa koordinatama unutar regiona, recimo (105,105), nakon
prebacivanja u relativne koordinate tačka bi trebala imati koorinate (5,5) unutar
samog regiona.
'''
region[y-min_y,x-min_x] = 255
regions_dict[min_x] = [resize_region(region), (min_x,min_y,max_x-min_x,max_y-min_y)]
sorted_regions_dict = collections.OrderedDict(sorted(regions_dict.items()))
sorted_regions = np.array(sorted_regions_dict.values())
return image_orig, sorted_regions[:, 0]
def sort_pixels(image,shape):
spill_pixels = []
outside_pixels = []
# test_image = np.zeros(image.shape)
# cv2.fillConvexPoly(test_image,shape,1)
border = 50
x,y,w,h = cv2.boundingRect(shape)
x = max(0,x-border)
y = max(0,y-border)
w += 2*border
h += 2*border
x2 = min(x+w,image.shape[1])
y2 = min(y+h,image.shape[0])
### Testing
# rect = test_image[y:y+h,x:x+w]
# cv2.imshow("test",rect)
# cv2.waitKey(0)
### Testing
for i in range(x,x2):
for j in range(y,y2):
# if test_image[j,i]:
dist = cv2.pointPolygonTest(shape,(i,j),True)
if dist >= 0:
spill_pixels.append(image[j,i])
elif dist >= -50:
outside_pixels.append(image[j,i])
# print cv2.contourArea(shape),len(pixels),cv2.arcLength(shape,True)
# assert cv2.contourArea(shape) == len(pixels)
return np.array(spill_pixels),np.array(outside_pixels)
def contains(self, point, closed=True):
"""do we contain point? if closed (in set-theoretic sense)
then we include points on our border, else False"""
contour = np.array([[x,y] for x,y in self.points], dtype=np.float32)
point = (np.float32(point.x()), np.float32(point.y()))
test = cv2.pointPolygonTest(contour, point, False)
return test > (-0.5 if closed else 0.5)
def distFromBorder(self, point):
"""
Returns the distance from the point to the border of the roi
:param tuple point: The (x, y) point to check
"""
return cv2.pointPolygonTest(self.points, point, True)
def minimumDistanceContour(contours, center):
distances = []
for i in range(0,len(contours)):
distance = cv2.pointPolygonTest(contours[i], center, True)
distances.append(np.fabs(distance))
index = distances.index(min(distances))
return index
def in_region(self, contour):
hits = 0
for (x, y) in self.control_pts:
if cv2.pointPolygonTest(contour, (x, y), False) >= 0:
hits += 1
# do majority vote of points
return hits > len(self.control_pts) / 2
def selectByPiont(self, row, col):
for c in self.contours:
if cv2.pointPolygonTest(c, (col,row), False) >= 0:
r = cv2.boundingRect(c)
i1, j1, i2, j2 = rect2coordinates(r)
return i1, j1, i2, j2
return -1, -1, -1, -1
def PolygonTestRC(contour, pt, count = 0, counter_code = ''):
count += 1
Inside = cv2.pointPolygonTest(contour, pt, False)
if Inside > -1:
counter_code = counter_code + '+' + str(count)
else:
counter_code = counter_code + '-' + str(count)
return count, counter_code
def in_contour(input_file, point, squares=True, thresh_val=255, k_size=5, iterations=30):
"""
>>> isinstance(in_contour(test_image, (200, 200)), np.ndarray)
True
>>> in_contour(test_image, (0, 0))
-1
>>> in_contour(None, (0,0))
Traceback (most recent call last):
File "<stdin>", line 1, in ?
IOError: The input file can't be a None object
>>> in_contour("", (0,0))
Traceback (most recent call last):
File "<stdin>", line 1, in ?
IOError: The input file can't be ''.
>>> in_contour("fakeRoute", (0,0))
Traceback (most recent call last):
File "<stdin>", line 1, in ?
IOError: Input file not found.
>>> in_contour(test_image, (0,0), thresh_val=-200)
Traceback (most recent call last):
File "<stdin>", line 1, in ?
ValueError: All threshold values must be between 0 and 255.
>>> in_contour(test_image, (0,0), k_size=-10)
Traceback (most recent call last):
File "<stdin>", line 1, in ?
ValueError: Kernel size value must be greater than 0.
>>> in_contour(test_image, (0,0), iterations=-10)
Traceback (most recent call last):
File "<stdin>", line 1, in ?
ValueError: Iterations value must be greater than 0.
"""
image = iu.get_image(input_file)
check_threshold(thresh_val)
check_kernel(k_size)
check_iterations(iterations)
contours = detect_contours(image, thresh_val, k_size, iterations)
if squares == True:
contours = get_squares(contours)
contours = join_contours(contours)
cnt = -1
for c in contours:
if cv.pointPolygonTest(c, point, measureDist=False) >= 0:
cnt = c
return cnt
def _get_palm_circle(contour, fore):
dist_max = np.zeros((fore.shape[0], fore.shape[1]))
for y in range(0, fore.shape[0], 4):
for x in range(0, fore.shape[1], 4):
if fore[y, x]:
dist_max[y, x] = cv2.pointPolygonTest(contour, (x, y), True)
min_val, max_val, min_loc, max_loc = cv2.minMaxLoc(dist_max)
return max_loc, max_val
def circleCheck(contours, hierarchy,circleX,circleY):
for h,cnt in enumerate(contours):
if cv2.pointPolygonTest(cnt,(circleX,circleY), False) == 1:
if hierarchy[0,h][2] == -1:
return h
else:
nuH = hierarchy[0,h][2]
kinderList = []
passed = True
if cv2.pointPolygonTest(contours[nuH],(circleX,circleY), False) != 1:
kinderList = recursiveNeighborCheck(kinderList,hierarchy,nuH,0)
kinderList = recursiveNeighborCheck(kinderList,hierarchy,nuH,1)
for kind in kinderList:
if cv2.pointPolygonTest(contours[kind],(circleX,circleY), False) == 1:
passed = False
if passed:
return h
return False
def testInside(self, p, measureDist=False):
"""tests to see if the point is inside the region described by
this Widget, or returns an approximate distance, depending on
whether measureDist is True."""
return cv2.pointPolygonTest(self.points, a2t(p), measureDist)
def isInContourV2(cont, pt, patch_size=256):
return 1 if cv2.pointPolygonTest(cont,
(pt[0] + patch_size / 2, pt[1] +
patch_size / 2), False) >= 0 else 0
def fix_hsv(x, y, param):
sliders = param[1].table
frame = param[0]
#print (sliders.value())
sliders['low h'].setValue(0)
sliders['low s'].setValue(100)
sliders['low v'].setValue(50)
sliders['high h'].setValue(187)
sliders['high s'].setValue(255)
sliders['high v'].setValue(255)
blur = cv2.GaussianBlur(frame, (15, 15), 0)
hsv = cv2.cvtColor(blur, cv2.COLOR_BGR2HSV)
sample = hsv[y, x]
print('clicked on', sample)
sliders['high h'].setValue(sample[0] + 1)
sliders['low h'].setValue(sample[0] - 1)
best_area = 0.0
best_radius = 0.0
object_found = False
while object_found == False:
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
for contour in cnts:
inside = cv2.pointPolygonTest(contour, (x, y), False)
if inside >= 0:
object_found = True
print('found')
((x, y), radius) = cv2.minEnclosingCircle(contour)
area = cv2.contourArea(contour)
best_area = area
break
print('inside:', inside)
if object_found == False:
sliders['high h'].setValue(sliders['high h'].value() + 1)
sliders['low h'].setValue(sliders['low h'].value() - 1)
print('best area', radius)
sliders['high h'].setValue(sliders['high h'].value() + 1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
for contour in cnts:
inside = cv2.pointPolygonTest(contour, (x, y), False)
if inside >= 0:
((x, y), radius) = cv2.minEnclosingCircle(contour)
area = cv2.contourArea(contour)
while area > best_area and radius > best_radius:
print('area', area, 'radius', radius)
best_radius = radius
best_area = area
sliders['high h'].setValue(sliders['high h'].value() + 1)
#sliders['low h'].setValue(sliders['low h'].value()-1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
for contour in cnts:
inside = cv2.pointPolygonTest(contour, (x, y), False)
if inside >= 0:
((x, y), radius) = cv2.minEnclosingCircle(contour)
area = cv2.contourArea(contour)
sliders['low h'].setValue(sliders['low h'].value() - 1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
for contour in cnts:
inside = cv2.pointPolygonTest(contour, (x, y), False)
if inside >= 0:
((x, y), radius) = cv2.minEnclosingCircle(contour)
area = cv2.contourArea(contour)
while area > best_area and radius > best_radius:
print('area', area, 'radius', radius)
best_radius = radius
best_area = area
#sliders['high h'].setValue(sliders['high h'].value()+1)
sliders['low h'].setValue(sliders['low h'].value() - 1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
for contour in cnts:
inside = cv2.pointPolygonTest(contour, (x, y), False)
if inside >= 0:
((x, y), radius) = cv2.minEnclosingCircle(contour)
area = cv2.contourArea(contour)
sliders['low s'].setValue(sliders['low s'].value() + 1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
while len(cnts) > 1:
sliders['low s'].setValue(sliders['low s'].value() + 1)
lower_red = np.array([
sliders['low h'].value(), sliders['low s'].value(),
sliders['low v'].value()
])
upper_red = np.array([
sliders['high h'].value(), sliders['high s'].value(),
sliders['high v'].value()
])
mask = cv2.inRange(hsv, lower_red, upper_red)
mask = cv2.erode(mask, None, iterations=2)
mask = cv2.dilate(mask, None, iterations=2)
cnts = cv2.findContours(mask.copy(), cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
cnts = cnts[0]
def analyze_object(img, obj, mask):
"""Outputs numeric properties for an input object (contour or grouped contours).
Inputs:
img = RGB or grayscale image data for plotting
obj = single or grouped contour object
mask = Binary image to use as mask for moments analysis
Returns:
shape_header = shape data table headers
shape_data = shape data table values
analysis_images = list of output images
:param img: numpy.ndarray
:param obj: list
:param mask: numpy.ndarray
:return shape_header: list
:return shape_data: list
:return analysis_images: list
"""
params.device += 1
# Valid objects can only be analyzed if they have >= 5 vertices
if len(obj) < 5:
return None, None, None
ori_img = np.copy(img)
# Convert grayscale images to color
if len(np.shape(ori_img)) == 2:
ori_img = cv2.cvtColor(ori_img, cv2.COLOR_GRAY2BGR)
if len(np.shape(img)) == 3:
ix, iy, iz = np.shape(img)
else:
ix, iy = np.shape(img)
size = ix, iy, 3
size1 = ix, iy
background = np.zeros(size, dtype=np.uint8)
background1 = np.zeros(size1, dtype=np.uint8)
background2 = np.zeros(size1, dtype=np.uint8)
# Check is object is touching image boundaries (QC)
frame_background = np.zeros(size1, dtype=np.uint8)
frame = frame_background + 1
frame_contour, frame_hierarchy = cv2.findContours(
frame, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
ptest = []
vobj = np.vstack(obj)
for i, c in enumerate(vobj):
xy = tuple(c)
pptest = cv2.pointPolygonTest(frame_contour[0], xy, measureDist=False)
ptest.append(pptest)
in_bounds = all(c == 1 for c in ptest)
# Convex Hull
hull = cv2.convexHull(obj)
hull_vertices = len(hull)
# Moments
# m = cv2.moments(obj)
m = cv2.moments(mask, binaryImage=True)
# Properties
# Area
area = m['m00']
if area:
# Convex Hull area
hull_area = cv2.contourArea(hull)
# Solidity
solidity = 1
if int(hull_area) != 0:
solidity = area / hull_area
# Perimeter
perimeter = cv2.arcLength(obj, closed=True)
# x and y position (bottom left?) and extent x (width) and extent y (height)
x, y, width, height = cv2.boundingRect(obj)
# Centroid (center of mass x, center of mass y)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
# Ellipse
center, axes, angle = cv2.fitEllipse(obj)
major_axis = np.argmax(axes)
minor_axis = 1 - major_axis
major_axis_length = axes[major_axis]
minor_axis_length = axes[minor_axis]
eccentricity = np.sqrt(1 - (axes[minor_axis] / axes[major_axis])**2)
# Longest Axis: line through center of mass and point on the convex hull that is furthest away
cv2.circle(background, (int(cmx), int(cmy)), 4, (255, 255, 255), -1)
center_p = cv2.cvtColor(background, cv2.COLOR_BGR2GRAY)
ret, centerp_binary = cv2.threshold(center_p, 0, 255,
cv2.THRESH_BINARY)
centerpoint, cpoint_h = cv2.findContours(centerp_binary, cv2.RETR_TREE,
cv2.CHAIN_APPROX_NONE)[-2:]
dist = []
vhull = np.vstack(hull)
for i, c in enumerate(vhull):
xy = tuple(c)
pptest = cv2.pointPolygonTest(centerpoint[0], xy, measureDist=True)
dist.append(pptest)
abs_dist = np.absolute(dist)
max_i = np.argmax(abs_dist)
caliper_max_x, caliper_max_y = list(tuple(vhull[max_i]))
caliper_mid_x, caliper_mid_y = [int(cmx), int(cmy)]
xdiff = float(caliper_max_x - caliper_mid_x)
ydiff = float(caliper_max_y - caliper_mid_y)
# Set default values
slope = 1
if xdiff != 0:
slope = (float(ydiff / xdiff))
b_line = caliper_mid_y - (slope * caliper_mid_x)
if slope != 0:
xintercept = int(-b_line / slope)
xintercept1 = int((ix - b_line) / slope)
if 0 <= xintercept <= iy and 0 <= xintercept1 <= iy:
cv2.line(background1, (xintercept1, ix), (xintercept, 0),
(255), params.line_thickness)
elif xintercept < 0 or xintercept > iy or xintercept1 < 0 or xintercept1 > iy:
# Used a random number generator to test if either of these cases were possible but neither is possible
# if xintercept < 0 and 0 <= xintercept1 <= iy:
# yintercept = int(b_line)
# cv2.line(background1, (0, yintercept), (xintercept1, ix), (255), 5)
# elif xintercept > iy and 0 <= xintercept1 <= iy:
# yintercept1 = int((slope * iy) + b_line)
# cv2.line(background1, (iy, yintercept1), (xintercept1, ix), (255), 5)
# elif 0 <= xintercept <= iy and xintercept1 < 0:
# yintercept = int(b_line)
# cv2.line(background1, (0, yintercept), (xintercept, 0), (255), 5)
# elif 0 <= xintercept <= iy and xintercept1 > iy:
# yintercept1 = int((slope * iy) + b_line)
# cv2.line(background1, (iy, yintercept1), (xintercept, 0), (255), 5)
# else:
yintercept = int(b_line)
yintercept1 = int((slope * iy) + b_line)
cv2.line(background1, (0, yintercept), (iy, yintercept1),
(255), 5)
else:
cv2.line(background1, (iy, caliper_mid_y), (0, caliper_mid_y),
(255), params.line_thickness)
ret1, line_binary = cv2.threshold(background1, 0, 255,
cv2.THRESH_BINARY)
# print_image(line_binary,(str(device)+'_caliperfit.png'))
cv2.drawContours(background2, [hull], -1, (255), -1)
ret2, hullp_binary = cv2.threshold(background2, 0, 255,
cv2.THRESH_BINARY)
# print_image(hullp_binary,(str(device)+'_hull.png'))
caliper = cv2.multiply(line_binary, hullp_binary)
# print_image(caliper,(str(device)+'_caliperlength.png'))
caliper_y, caliper_x = np.array(caliper.nonzero())
caliper_matrix = np.vstack((caliper_x, caliper_y))
caliper_transpose = np.transpose(caliper_matrix)
caliper_length = len(caliper_transpose)
caliper_transpose1 = np.lexsort((caliper_y, caliper_x))
caliper_transpose2 = [(caliper_x[i], caliper_y[i])
for i in caliper_transpose1]
caliper_transpose = np.array(caliper_transpose2)
# else:
# hull_area, solidity, perimeter, width, height, cmx, cmy = 'ND', 'ND', 'ND', 'ND', 'ND', 'ND', 'ND'
# Store Shape Data
shape_header = [
'HEADER_SHAPES', 'area', 'hull-area', 'solidity', 'perimeter', 'width',
'height', 'longest_axis', 'center-of-mass-x', 'center-of-mass-y',
'hull_vertices', 'in_bounds', 'ellipse_center_x', 'ellipse_center_y',
'ellipse_major_axis', 'ellipse_minor_axis', 'ellipse_angle',
'ellipse_eccentricity'
]
shape_data = [
'SHAPES_DATA', area, hull_area, solidity, perimeter, width, height,
caliper_length, cmx, cmy, hull_vertices, in_bounds, center[0],
center[1], major_axis_length, minor_axis_length, angle, eccentricity
]
analysis_images = []
# Draw properties
if area:
cv2.drawContours(ori_img, obj, -1, (255, 0, 0), params.line_thickness)
cv2.drawContours(ori_img, [hull], -1, (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (x, y), (x + width, y), (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height), (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (tuple(caliper_transpose[caliper_length - 1])),
(tuple(caliper_transpose[0])), (0, 0, 255),
params.line_thickness)
cv2.circle(ori_img, (int(cmx), int(cmy)), 10, (0, 0, 255),
params.line_thickness)
# Output images with convex hull, extent x and y
# out_file = os.path.splitext(filename)[0] + '_shapes.jpg'
# out_file1 = os.path.splitext(filename)[0] + '_mask.jpg'
# print_image(ori_img, out_file)
analysis_images.append(ori_img)
# print_image(mask, out_file1)
analysis_images.append(mask)
else:
pass
# Store into global measurements
if not "shapes" in outputs.measurements:
outputs.measurements["shapes"] = {}
outputs.measurements["shapes"]["area"] = area
outputs.measurements["shapes"]["hull-area"] = hull_area
outputs.measurements["shapes"]["solidity"] = solidity
outputs.measurements["shapes"]["perimeter"] = perimeter
outputs.measurements["shapes"]["width"] = width
outputs.measurements["shapes"]["height"] = height
outputs.measurements["shapes"]["longest_axis"] = caliper_length
outputs.measurements["shapes"]["center-of-mass-x"] = cmx
outputs.measurements["shapes"]["center-of-mass-y"] = cmy
outputs.measurements["shapes"]["hull_vertices"] = hull_vertices
outputs.measurements["shapes"]["in_bounds"] = in_bounds
outputs.measurements["shapes"]["ellipse_center_x"] = center[0]
outputs.measurements["shapes"]["ellipse_center_y"] = center[1]
outputs.measurements["shapes"]["ellipse_major_axis"] = major_axis_length
outputs.measurements["shapes"]["ellipse_minor_axis"] = minor_axis_length
outputs.measurements["shapes"]["ellipse_angle"] = angle
outputs.measurements["shapes"]["ellipse_eccentricity"] = eccentricity
if params.debug is not None:
cv2.drawContours(ori_img, obj, -1, (255, 0, 0), params.line_thickness)
cv2.drawContours(ori_img, [hull], -1, (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (x, y), (x + width, y), (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height), (0, 0, 255),
params.line_thickness)
cv2.circle(ori_img, (int(cmx), int(cmy)), 10, (0, 0, 255),
params.line_thickness)
cv2.line(ori_img, (tuple(caliper_transpose[caliper_length - 1])),
(tuple(caliper_transpose[0])), (0, 0, 255),
params.line_thickness)
if params.debug == 'print':
print_image(
ori_img,
os.path.join(params.debug_outdir,
str(params.device) + '_shapes.jpg'))
elif params.debug == 'plot':
if len(np.shape(img)) == 3:
plot_image(ori_img)
else:
plot_image(ori_img, cmap='gray')
# Store images
outputs.images.append(analysis_images)
return shape_header, shape_data, analysis_images
def __call__(self, pt):
return 1 if cv2.pointPolygonTest(
self.cont, (pt[0] + self.patch_size // 2,
pt[1] + self.patch_size // 2), False) >= 0 else 0
def findDefects():
cap = cv2.VideoCapture(0)
kernel=(5,5)
if(cap.isOpened()==False):
print('Unable to read camera feed')
count=0
global cx;global cy;
while(True):
ret, frame = cap.read()
# frame=cv2.resize(frame,(1920,1080))
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
blur = cv2.bilateralFilter(gray,10,50,50)
ret,thresh1 = cv2.threshold(blur,110,255,cv2.THRESH_BINARY_INV+cv2.THRESH_OTSU)
thresh1 = cv2.morphologyEx(thresh1, cv2.MORPH_OPEN, kernel)
thresh1 = cv2.morphologyEx(thresh1, cv2.MORPH_CLOSE, kernel)
blah,contours,hierarchy = cv2.findContours(thresh1,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
# blah,contours = cv2.findContours(thresh1,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
max_area=0;ci=0;
for i in range(len(contours)):
cnt=contours[i]
area = cv2.contourArea(cnt)
if(area>max_area):
max_area=area
ci=i
cnt = contours[ci]
hull = cv2.convexHull(cnt)
moments = cv2.moments(cnt)
if moments['m00']!=0:
cx = int(moments['m10']/moments['m00'])
cy = int(moments['m01']/moments['m00'])
#print(cx,cy,end='\r')
center = (cx,cy)
# pyautogui.moveTo(cx,cy)
cnt = cv2.approxPolyDP(cnt,0.01*cv2.arcLength(cnt,True),True)
drawing = np.zeros(frame.shape,np.uint8)
frame = cv2.circle(drawing,center,5,[0,255,255],2)
frame = cv2.drawContours(drawing,[cnt],0,(0,255,0),2)
frame = cv2.drawContours(drawing,[hull],0,(0,0,255),2)
hull = cv2.convexHull(cnt,returnPoints = False)
defects = cv2.convexityDefects(cnt,hull)
if defects is None:
continue
mind=0;maxd=0;i=0;fin=0;
for i in range(defects.shape[0]):
s,e,f,d = defects[i,0]
start = tuple(cnt[s][0])
end = tuple(cnt[e][0])
far = tuple(cnt[f][0])
dist = cv2.pointPolygonTest(cnt,center,True)
frame = cv2.line(frame,start,end,[255,0,0],2)
frame = cv2.circle(frame,far,5,[0,0,255],-1)
a = math.sqrt((end[0] - start[0]) ** 2 + (end[1] - start[1]) ** 2)
b = math.sqrt((far[0] - start[0]) ** 2 + (far[1] - start[1]) ** 2)
c = math.sqrt((end[0] - far[0]) ** 2 + (end[1] - far[1]) ** 2)
angle = math.acos((b ** 2 + c ** 2 - a ** 2) / (2 * b * c))
if angle <= math.pi / 2:
fin+=1
#cv2.circle(drawing, far, 8, [211, 84, 0], -1)
#print('fingers=',fin)
cv2.imshow('frame',frame)
if cv2.waitKey(5) & 0xFF == 27:
break
if cv2.waitKey(25) & 0xFF == ord('q'):
cv2.destroyAllWindows()
break
if(fin<=3):
count=count+1
clicker(count)
cap.release()
cv2.destroyAllWindows()
def analyze_bound_vertical(img, obj, mask, line_position, filename=False):
"""User-input boundary line tool
Inputs:
img = RGB or grayscale image data for plotting
obj = single or grouped contour object
mask = Binary mask made from selected contours
shape_header = pass shape header data to function
shape_data = pass shape data so that analyze_bound data can be appended to it
line_position = position of boundry line (a value of 0 would draw the line through the left side of the image)
filename = False or image name. If defined print image.
Returns:
bound_header = data table column headers
bound_data = boundary data table
analysis_images = output image filenames
:param img: numpy.ndarray
:param obj: list
:param mask: numpy.ndarray
:param line_position: int
:param filename: str
:return bound_header: tuple
:return bound_data: tuple
:return analysis_images: list
"""
params.device += 1
ori_img = np.copy(img)
# Draw line horizontal line through bottom of image, that is adjusted to user input height
if len(np.shape(ori_img)) == 2:
ori_img = cv2.cvtColor(ori_img, cv2.COLOR_GRAY2BGR)
iy, ix, iz = np.shape(ori_img)
size = (iy, ix)
size1 = (iy, ix, 3)
background = np.zeros(size, dtype=np.uint8)
wback = (np.zeros(size1, dtype=np.uint8)) + 255
x_coor = 0 + int(line_position)
y_coor = int(iy)
rec_point1 = (0, 0)
rec_point2 = (x_coor, y_coor - 2)
cv2.rectangle(background, rec_point1, rec_point2, (255), -1)
right_contour, right_hierarchy = cv2.findContours(background, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
x, y, width, height = cv2.boundingRect(obj)
if x_coor - x <= 0:
width_left_bound = 0
width_right_bound = width
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
width_left_bound = width
width_right_bound = 0
else:
width_left_bound = x_coor - x
width_right_bound = width - width_left_bound
right = []
left = []
mask_nonzerox, mask_nonzeroy = np.nonzero(mask)
obj_points = np.vstack((mask_nonzeroy, mask_nonzerox))
obj_points1 = np.transpose(obj_points)
for i, c in enumerate(obj_points1):
xy = tuple(c)
pptest = cv2.pointPolygonTest(right_contour[0], xy, measureDist=False)
if pptest == 1:
left.append(xy)
cv2.circle(ori_img, xy, 1, (0, 0, 255))
cv2.circle(wback, xy, 1, (0, 0, 255))
else:
right.append(xy)
cv2.circle(ori_img, xy, 1, (0, 255, 0))
cv2.circle(wback, xy, 1, (0, 255, 0))
right_bound_area = len(right)
left_bound_area = len(left)
percent_bound_area_right = ((float(right_bound_area)) / (float(left_bound_area + right_bound_area))) * 100
percent_bound_area_left = ((float(left_bound_area)) / (float(right_bound_area + left_bound_area))) * 100
bound_header = [
'HEADER_BOUNDARY' + str(line_position),
'width_left_bound',
'width_right_bound',
'left_bound_area',
'percent_left_bound_area',
'right_bound_area',
'percent_right_bound_area'
]
bound_data = [
'BOUNDARY_DATA',
width_left_bound,
width_right_bound,
left_bound_area,
percent_bound_area_left,
right_bound_area,
percent_bound_area_right
]
analysis_images = []
if left_bound_area or right_bound_area:
point3 = (x_coor+2, 0)
point4 = (x_coor+2, y_coor)
cv2.line(ori_img, point3, point4, (255, 0, 255), 5)
cv2.line(wback, point3, point4, (255, 0, 255), 5)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if x_coor - x <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0), 3)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0), 3)
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)), (255, 0, 0), 3)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (255, 0, 0), 3)
else:
cv2.line(ori_img, (x_coor + 2, int(cmy)), (x_coor + width_left_bound, int(cmy)), (255, 0, 0), 3)
cv2.line(ori_img, (x_coor + 2, int(cmy)), (x_coor - width_right_bound, int(cmy)), (0, 255, 0), 3)
cv2.line(wback, (x_coor + 2, int(cmy)), (x_coor + width_left_bound, int(cmy)), (255, 0, 0), 3)
cv2.line(wback, (x_coor + 2, int(cmy)), (x_coor - width_right_bound, int(cmy)), (0, 255, 0), 3)
if filename:
# Output images with boundary line, above/below bound area
out_file = str(filename[0:-4]) + '_boundary' + str(line_position) + '.jpg'
print_image(ori_img, out_file)
analysis_images = ['IMAGE', 'boundary', out_file]
if params.debug is not None:
point3 = (x_coor+2, 0)
point4 = (x_coor+2, y_coor)
cv2.line(ori_img, point3, point4, (255, 0, 255), 5)
cv2.line(wback, point3, point4, (255, 0, 255), 5)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if x_coor - x <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0), 3)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0), 3)
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)), (255, 0, 0), 3)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (255, 0, 0), 3)
else:
cv2.line(ori_img, (x_coor + 2, int(cmy)), (x_coor + width_left_bound, int(cmy)), (255, 0, 0), 3)
cv2.line(ori_img, (x_coor + 2, int(cmy)), (x_coor - width_right_bound, int(cmy)), (0, 255, 0), 3)
cv2.line(wback, (x_coor + 2, int(cmy)), (x_coor + width_left_bound, int(cmy)), (255, 0, 0), 3)
cv2.line(wback, (x_coor + 2, int(cmy)), (x_coor - width_right_bound, int(cmy)), (0, 255, 0), 3)
if params.debug == 'print':
print_image(wback, os.path.join(params.debug_outdir, str(params.device) + '_boundary_on_white.jpg'))
print_image(ori_img, os.path.join(params.debug_outdir, str(params.device) + '_boundary_on_img.jpg'))
if params.debug == 'plot':
plot_image(wback)
plot_image(ori_img)
return bound_header, bound_data, analysis_images
def check_modular_boxes(self):
if len(self.contour_to_show) <= 0:
self.OrientationStatus = "BAD"
return self.OrientationStatus
BlockRectanglesGood = []
BlockRectanglesFlip = []
TestResultsGood = []
TestResultsFlip = []
img_box = cv2.cvtColor(self.image, cv2.COLOR_GRAY2BGR)
self.OrientationStatus = "BAD"
#self.box = cv2.minAreaRect(self.contour_to_show)
box_angle = self.box[2]
box_dimensions = self.box[1]
box = cv2.cv.BoxPoints(self.box)
box = np.array(box, dtype="int")
cv2.drawContours(img_box, [box], -1, (255, 0, 0), 2)
cv2.drawContours(img_box, self.contour_to_show, -1, (0, 255, 0), 2)
if box_dimensions[0] < box_dimensions[1]:
box_angle += 90
#print("box angle: ",box_angle)
#BlockRectanglesGood=np.zeros((1,4,2), dtype=np.int)
if box_angle < 0:
boxSorted = [
box[box[:, 1].argmin()], box[box[:, 0].argmax()],
box[box[:, 1].argmax()], box[box[:, 0].argmin()]
]
boxTopRightX = boxSorted[0][0]
boxTopRightY = boxSorted[0][1]
boxBottomRightX = boxSorted[1][0]
boxBottomRightY = boxSorted[1][1]
boxBottomLeftX = boxSorted[2][0]
boxBottomLeftY = boxSorted[2][1]
boxTopLeftX = boxSorted[3][0]
boxTopLeftY = boxSorted[3][1]
elif box_angle > 0:
boxSorted = [
box[box[:, 1].argmin()], box[box[:, 0].argmax()],
box[box[:, 1].argmax()], box[box[:, 0].argmin()]
]
boxTopLeftX = boxSorted[0][0]
boxTopLeftY = boxSorted[0][1]
boxTopRightX = boxSorted[1][0]
boxTopRightY = boxSorted[1][1]
boxBottomRightX = boxSorted[2][0]
boxBottomRightY = boxSorted[2][1]
boxBottomLeftX = boxSorted[3][0]
boxBottomLeftY = boxSorted[3][1]
else:
maxX = box[box[:, 0].argmax()][0]
maxY = box[box[:, 1].argmax()][1]
#boxSorted=[box[box[:,1].argmin()],box[box[:,0].argmax()],box[box[:,1].argmax()],box[box[:,0].argmin()]]
for i in range(0, 4):
if box[i][0] == maxX:
if box[i][1] == maxY:
boxBottomRightX = box[i][0]
boxBottomRightY = box[i][1]
#print("bottomRx: "+str(boxBottomRightX)+" bottomRy: "+str(boxBottomRightY))
else:
boxTopRightX = box[i][0]
boxTopRightY = box[i][1]
#print("TopRightx: "+str(boxTopRightX)+" TopRighty: "+str(boxTopRightY))
else:
if box[i][1] == maxY:
boxBottomLeftX = box[i][0]
boxBottomLeftY = box[i][1]
#print("bottomLx: "+str(boxBottomLeftX)+" bottomLy: "+str(boxBottomLeftY))
else:
boxTopLeftX = box[i][0]
boxTopLeftY = box[i][1]
#print("TopLeftx: "+str(boxTopLeftX)+" TopLefty: "+str(boxTopLeftY))
boxHeightX = boxBottomLeftX - boxTopLeftX
boxHeightY = boxBottomLeftY - boxTopLeftY
boxWidthX = boxBottomRightX - boxBottomLeftX
boxWidthY = boxBottomRightY - boxBottomLeftY
#print("box: ",box)
#print("boxWidthX :"+str(boxWidthX)+" boxWidthY :"+str(boxWidthY)+" boxHeightX :"+str(boxHeightX)+" boxHeightY :"+str(boxHeightY)+" boxTopLeftX :"+str(boxTopLeftX)+" boxTopLeftY :"+str(boxTopLeftY))
for i in range(0, 10):
#print("EmptyBoxes: ",self.EmptyBoxes[i])
if sum(self.EmptyBoxes[i]) > 0:
#print("we draw rectangle i: ",i)
#print("EmptyBoxes in if: ",self.EmptyBoxes[i])
#print("BlockRect: ",BlockRectanglesGood)
TopLeftX = boxTopLeftX + int(
self.EmptyBoxes[i][0] * boxWidthX) + int(
self.EmptyBoxes[i][1] * boxHeightX)
TopLeftY = boxTopLeftY + int(
self.EmptyBoxes[i][0] * boxWidthY) + int(
self.EmptyBoxes[i][1] * boxHeightY)
#print("WidthY :"+str(self.EmptyBoxes[i][1]*boxWidthY)+" HeightY: "+str(self.EmptyBoxes[i][1]*boxHeightY))
TopRightX = TopLeftX + int(self.EmptyBoxes[i][2] * boxWidthX)
TopRightY = TopLeftY + int(self.EmptyBoxes[i][2] * boxWidthY)
BottomRightX = TopRightX + int(
self.EmptyBoxes[i][3] * boxHeightX)
BottomRightY = TopRightY + int(
self.EmptyBoxes[i][3] * boxHeightY)
BottomLeftX = BottomRightX - int(
self.EmptyBoxes[i][2] * boxWidthX)
BottomLeftY = BottomRightY - int(
self.EmptyBoxes[i][2] * boxWidthY)
TopLeft = [TopLeftX, TopLeftY]
TopRight = [TopRightX, TopRightY]
BottomLeft = [BottomLeftX, BottomLeftY]
BottomRight = [BottomRightX, BottomRightY]
to_append = [TopLeft, TopRight, BottomRight, BottomLeft]
to_append = np.array(to_append, dtype="int")
BlockRectanglesGood.append(to_append)
BottomRightX = boxBottomRightX - int(
self.EmptyBoxes[i][0] * boxWidthX) - int(
self.EmptyBoxes[i][1] * boxHeightX)
BottomRightY = boxBottomRightY - int(
self.EmptyBoxes[i][0] * boxWidthY) - int(
self.EmptyBoxes[i][1] * boxHeightY)
BottomLeftX = BottomRightX - int(
self.EmptyBoxes[i][2] * boxWidthX)
BottomLeftY = BottomRightY - int(
self.EmptyBoxes[i][2] * boxWidthY)
TopLeftX = BottomLeftX - int(
self.EmptyBoxes[i][3] * boxHeightX)
TopLeftY = BottomLeftY - int(
self.EmptyBoxes[i][3] * boxHeightY)
TopRightX = TopLeftX + int(self.EmptyBoxes[i][2] * boxWidthX)
TopRightY = TopLeftY + int(self.EmptyBoxes[i][2] * boxWidthY)
TopLeft = [TopLeftX, TopLeftY]
TopRight = [TopRightX, TopRightY]
BottomLeft = [BottomLeftX, BottomLeftY]
BottomRight = [BottomRightX, BottomRightY]
to_append = [TopLeft, TopRight, BottomRight, BottomLeft]
to_append = np.array(to_append, dtype="int")
BlockRectanglesFlip.append(to_append)
#BlockRectanglesGood = np.array(BlockRectanglesGood, dtype="int")
#print("we draw rectangle points: ",len(BlockRectanglesFlip))
statusText = ""
for i in range(0, len(self.contour_to_show)):
contourPoint = (self.contour_to_show[i][0][0],
self.contour_to_show[i][0][1])
for k in range(0, len(BlockRectanglesFlip)):
cv2.drawContours(img_box, [BlockRectanglesGood[k]], -1,
(255, 255, 0), 2) #light blue
cv2.drawContours(img_box, [BlockRectanglesFlip[k]], -1,
(0, 255, 255), 2) #yellow
if len(TestResultsGood) < k + 1:
TestResultsGood.append(
cv2.pointPolygonTest(BlockRectanglesGood[k],
contourPoint, False))
else:
if TestResultsGood[k] <= 0:
TestResultsGood[k] = cv2.pointPolygonTest(
BlockRectanglesGood[k], contourPoint, False)
if len(TestResultsFlip) < k + 1:
TestResultsFlip.append(
cv2.pointPolygonTest(BlockRectanglesFlip[k],
contourPoint, False))
else:
if TestResultsFlip[k] <= 0:
TestResultsFlip[k] = cv2.pointPolygonTest(
BlockRectanglesFlip[k], contourPoint, False)
#print("TestResultsGood: ",max(TestResultsGood))
if max(TestResultsGood) < 0 and max(TestResultsFlip) > 0:
self.OrientationStatus = "GOOD"
elif max(TestResultsGood) > 0 and max(TestResultsFlip) < 0:
self.OrientationStatus = "FLIP"
else:
self.OrientationStatus = "BAD"
cv2.putText(
img_box, " Box Dimensions: " + str(box_dimensions) + " Time: " +
str(time.time()), (10, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.45,
(255, 255, 255), 2)
#cv2.putText(img_box," Box Dimensions: "+str(box_dimensions)+" Time: "+str(time.time())+" Delta Time: "+str(QualityMeasures['deltaTime'])+" Delta Good: "+str(QualityMeasures['deltaGood']),(10,30),cv2.FONT_HERSHEY_SIMPLEX,0.45, (255, 255, 255), 2)
cv2.putText(
img_box, "ORIENTATION: " + self.OrientationStatus + " SIZE: " +
self.SizeStatus, (10, self.dimX - 30), cv2.FONT_HERSHEY_SIMPLEX, 1,
(255, 255, 255), 2)
#print("TestResultsFlip: ",TestResultsFlip)
#if len(TestResultsGood)
self.analyzedImage = img_box
return self.OrientationStatus
cy = int(moments['m01'] / moments['m00']) # cy = M01/M00
centr = (cx, cy)
cv2.circle(img, centr, 5, [0, 0, 255], 2)
cv2.drawContours(drawing, [cnt], 0, (0, 255, 0), 2)
cv2.drawContours(drawing, [hull], 0, (0, 0, 255), 2)
cnt = cv2.approxPolyDP(cnt, 0.01 * cv2.arcLength(cnt, True), True)
hull = cv2.convexHull(cnt, returnPoints=False)
if (1):
defects = cv2.convexityDefects(cnt, hull)
mind = 0
maxd = 0
for i in range(defects.shape[0]):
s, e, f, d = defects[i, 0]
start = tuple(cnt[s][0])
end = tuple(cnt[e][0])
far = tuple(cnt[f][0])
dist = cv2.pointPolygonTest(cnt, centr, True)
cv2.line(img, start, end, [0, 255, 0], 2)
cv2.circle(img, far, 5, [0, 0, 255], -1)
print(i)
i = 0
cv2.imshow('output', drawing)
cv2.imshow('input', img)
k = cv2.waitKey(10)
if k == 27:
break
topx = 0
topy = 0
botx = 100000000000000
boty = 100000000000000
for a in range(0, len(c)):
if (c[a][0][0] > topx):
topx = c[a][0][0]
if (c[a][0][1] > topy):
topy = c[a][0][1]
if (c[a][0][0] < botx):
botx = c[a][0][0]
if (c[a][0][1] < boty):
boty = c[a][0][1]
print(topx, topy, botx, boty)
for b in range(1, len(discs) + 1):
dist = cv2.pointPolygonTest(c, (discs[b]), False)
if (dist > 0):
inzone.append(b)
print(discs)
matchdiscs = matchdiscs + len(inzone)
if (len(inzone) == 1):
minzonedist = 10000000000000000000000000000000000
mindishdist = minzonedist
radii = []
avgradius = 0
for f in range(0, len(c)):
xdist = c[f][0][0] - discs[inzone[0]][0]
ydist = c[f][0][1] - discs[inzone[0]][1]
def contains_point(self, point):
"""Is the provided point inside of the Polygon?"""
return cv2.pointPolygonTest(self.points, point, False) > 0
def analyze_bound_horizontal(img, obj, mask, line_position, label="default"):
"""User-input boundary line tool
Inputs:
img = RGB or grayscale image data for plotting
obj = single or grouped contour object
mask = Binary mask made from selected contours
line_position = position of boundary line (a value of 0 would draw the line through the top of the image)
label = optional label parameter, modifies the variable name of observations recorded
Returns:
analysis_images = list of output images
:param img: numpy.ndarray
:param obj: list
:param mask: numpy.ndarray
:param line_position: int
:param label: str
:return analysis_images: list
"""
ori_img = np.copy(img)
# Draw line horizontal line through bottom of image, that is adjusted to user input height
if len(np.shape(ori_img)) == 2:
ori_img = cv2.cvtColor(ori_img, cv2.COLOR_GRAY2BGR)
iy, ix, iz = np.shape(ori_img)
size = (iy, ix)
size1 = (iy, ix, 3)
background = np.zeros(size, dtype=np.uint8)
wback = (np.zeros(size1, dtype=np.uint8)) + 255
x_coor = int(ix)
y_coor = line_position
rec_corner = int(iy - 2)
rec_point1 = (1, rec_corner)
rec_point2 = (x_coor - 2, y_coor - 2)
cv2.rectangle(background, rec_point1, rec_point2, (255), 1)
below_contour, below_hierarchy = cv2.findContours(
background, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
x, y, width, height = cv2.boundingRect(obj)
if y_coor - y <= 0:
height_above_bound = 0
height_below_bound = height
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
height_above_bound = height
height_below_bound = 0
else:
height_above_bound = y_coor - y
height_below_bound = height - height_above_bound
below = []
above = []
mask_nonzerox, mask_nonzeroy = np.nonzero(mask)
obj_points = np.vstack((mask_nonzeroy, mask_nonzerox))
obj_points1 = np.transpose(obj_points)
for i, c in enumerate(obj_points1):
xy = tuple(c)
pptest = cv2.pointPolygonTest(below_contour[0], xy, measureDist=False)
if pptest == 1:
below.append(xy)
cv2.circle(ori_img, xy, 1, (155, 0, 255))
cv2.circle(wback, xy, 1, (155, 0, 255))
else:
above.append(xy)
cv2.circle(ori_img, xy, 1, (0, 255, 0))
cv2.circle(wback, xy, 1, (0, 255, 0))
above_bound_area = len(above)
below_bound_area = len(below)
percent_bound_area_above = (
(float(above_bound_area)) /
(float(above_bound_area + below_bound_area))) * 100
percent_bound_area_below = (
(float(below_bound_area)) /
(float(above_bound_area + below_bound_area))) * 100
analysis_images = []
if above_bound_area or below_bound_area:
point3 = (0, y_coor - 4)
point4 = (x_coor, y_coor - 4)
cv2.line(ori_img, point3, point4, (255, 0, 255), params.line_thickness)
cv2.line(wback, point3, point4, (255, 0, 255), params.line_thickness)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if y_coor - y <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(0, 255, 0), params.line_thickness)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height), (0, 255, 0),
params.line_thickness)
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), params.line_thickness)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), params.line_thickness)
else:
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
params.line_thickness)
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
params.line_thickness)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
params.line_thickness)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
params.line_thickness)
# Output image with boundary line, above/below bound area
analysis_images.append(wback)
analysis_images.append(ori_img)
if params.debug is not None:
params.device += 1
point3 = (0, y_coor - 4)
point4 = (x_coor, y_coor - 4)
cv2.line(ori_img, point3, point4, (255, 0, 255), params.line_thickness)
cv2.line(wback, point3, point4, (255, 0, 255), params.line_thickness)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if y_coor - y <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(0, 255, 0), params.line_thickness)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height), (0, 255, 0),
params.line_thickness)
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), params.line_thickness)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), params.line_thickness)
else:
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
params.line_thickness)
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
params.line_thickness)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
params.line_thickness)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
params.line_thickness)
if params.debug == 'print':
print_image(
wback,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_white.png'))
print_image(
ori_img,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_img.png'))
if params.debug == 'plot':
plot_image(wback)
plot_image(ori_img)
outputs.add_observation(sample=label,
variable='horizontal_reference_position',
trait='horizontal reference position',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='none',
datatype=int,
value=line_position,
label='none')
outputs.add_observation(sample=label,
variable='height_above_reference',
trait='height above reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='pixels',
datatype=int,
value=height_above_bound,
label='pixels')
outputs.add_observation(sample=label,
variable='height_below_reference',
trait='height_below_reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='pixels',
datatype=int,
value=height_below_bound,
label='pixels')
outputs.add_observation(sample=label,
variable='area_above_reference',
trait='area above reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='pixels',
datatype=int,
value=above_bound_area,
label='pixels')
outputs.add_observation(sample=label,
variable='percent_area_above_reference',
trait='percent area above reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='none',
datatype=float,
value=percent_bound_area_above,
label='none')
outputs.add_observation(sample=label,
variable='area_below_reference',
trait='area below reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='pixels',
datatype=int,
value=below_bound_area,
label='pixels')
outputs.add_observation(sample=label,
variable='percent_area_below_reference',
trait='percent area below reference',
method='plantcv.plantcv.analyze_bound_horizontal',
scale='none',
datatype=float,
value=percent_bound_area_below,
label='none')
# Store images
outputs.images.append(analysis_images)
return ori_img
cv2.circle(img, tuple(cnt[cnt[:, :, 1].argmin()][0]), 2, (0, 0, 255), 2)
cv2.circle(img, tuple(cnt[cnt[:, :, 1].argmax()][0]), 2, (0, 0, 255), 2)
cv2.imshow('img', img)
# region
hull = cv2.convexHull(cnt, returnPoints=False)
defects = cv2.convexityDefects(cnt, hull)
for i in range(defects.shape[0]):
s, e, f, d = defects[i, 0]
# cv2.circle(img, tuple(cnt[s][0]), 2, (0, 0, 255), 2)
# cv2.circle(img, tuple(cnt[e][0]), 2, (0, 255, 0), 2)
cv2.line(img, tuple(cnt[s][0]), tuple(cnt[e][0]), (0, 0, 255), 1)
cv2.circle(img, tuple(cnt[f][0]), 2, (0, 255, 0), 2)
cv2.imshow('img', img)
dist = cv2.pointPolygonTest(cnt, (50, 50), True)
inner = cv2.pointPolygonTest(cnt, (343, 218), False)
img = cv2.imread('f:/index.jpg')
img = cv2.drawContours(img, contours, 7, (0, 0, 255), 2)
img = cv2.drawContours(img, contours, 9, (0, 0, 255), 2)
cv2.imshow('img', img)
cnt1 = contours[7]
cnt2 = contours[9]
cv2.matchShapes(cnt1, cnt2, cv2.CONTOURS_MATCH_I1, 0.0)
# retrieval
image, contours, hierarchy = cv2.findContours(th, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_SIMPLE)
# histogram
def imgProcess(img):
# NEED TO SAVE AS SAME THING EACH TIME, 0 converts to grayscale
# img = cv2.imread('test.jpg', 0)
# Resize image 300 x 225 pixels params
r = 400.0 / img.shape[1]
dim = (400, int(img.shape[0] * r))
# perform the actual resizing of the image and show it
kSize = 3
kernel = np.ones((kSize, kSize), np.uint8)
resized = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# smoothed = cv2.filter2D(img, -1, kernel) # Smooth image with 15x15 array
opening = cv2.morphologyEx(resized, cv2.MORPH_CLOSE, kernel)
# Resize image
# cv2.imwrite('Resized.jpg', resized)
# apply automatic Canny edge detection using the computed median
# Define automatic upper and lower thresholds for Canny detection
v = np.median(opening)
sigma = 0.333
lower = int(max(0, (1.0 - sigma) * v))
upper = int(min(255, (1.0 + sigma) * v))
edges = cv2.Canny(opening, lower, upper) # Perform Canny Edge Detection
# edge_save = cv2.imwrite('edged.jpg', edges)
# Resized colour version of original image
smallColr = cv2.resize(img, dim, interpolation=cv2.INTER_AREA)
# cv2.imshow('Edges', edges)
# Find contours in canny image
image, contours, hierarchy = cv2.findContours(edges, cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
showcontours = cv2.drawContours(smallColr, contours, -1, (0, 255, 0), 3)
#cv2.imwrite('origcontours.jpg', showcontours)
#cv2.imshow('original contours',showcontours)
# for i in range(len(contours)):
# color = np.random.rand(3) * 255
# cnt_img = cv2.drawContours(cnt_img, contours, i, color, 3)
from perspCorr2 import perspMatrix
h = perspMatrix # import perspective distortion matrix from other file
# cnt_img is transformed image (transform small colour image)
# cnt_img = cv2.warpPerspective(smallColr, h, (4000, 3000))
# cv2.imshow('Warped', cnt_img)
# cv2.imshow('Original', smallColr)
# Added finding contour work:
# This assumes that the robot is already on the line
# create array containing closest contours to robot
# duplicate contours list and remove any contours that are too short
contours_shorten = list(contours)
minLength = 19
# remove any small contours (typically errors)
for cont in range(len(contours)):
if len(contours[cont]) < minLength:
contours_shorten.remove(contours[cont])
contour_pertr = [] # perspective correction contours
contours_pertrInts = []
# iterate through contour list
for cont3 in range(len(contours_shorten)):
contours_shorten[cont3] = contours_shorten[cont3].reshape(-1, 2)
# take each contour in term
a = contours_shorten[cont3]
a = np.array([a])
a = a.astype(float)
# apply perspective transform to each contour
contour_pertr.append(cv2.perspectiveTransform(a, h))
contour_pertr[cont3] = np.reshape(np.ravel(contour_pertr[cont3]),
(-1, 2))
contours_pertrInts.append(cv2.perspectiveTransform(a, h))
contours_pertrInts[cont3] = np.reshape(
np.ravel(contours_pertrInts[cont3]), (-1, 2))
# contours_pertrInts[cont3] = contours_pertrInts[cont3].astype(int)
# for ind in range(2):
# contours_pertrInts[cont3][cont_pertr][ind] = int(contour_pertr[cont3][cont_pertr][ind])
# contours_pertrInts[cont3] = np.reshape(np.ravel(contours_pertrInts[cont3]), (-1, 1, 2))
# contours_pertrInts[cont3].astype(int)
# contour_pertr[cont3] = np.reshape(np.ravel(contour_pertr[cont3]), (-1, 1, 2))
# contours_shorten[cont3] = a.reshape(-1, 1, 2)
# contours = contours3
# contour_pertr.astype(int)
# draw contours to image
# cnt_img2 = cv2.drawContours(cnt_img, contours_pertrInts, -1, (0, 255, 0), 3)
# cv2.imshow('Contours', cnt_img2)
closestContours = []
for i in range(len(contours)): # iterate through each contour
for point in range(len(
contours[i])): # iterate through the points in each contour
# find contours with points in middle 165-235 pixels x and bottom 280-300 y - i.e. closest contours
# thresholds can be changed to ensure just 2 contours are found
if 100 <= contours[i][point][0][0] <= 300 and 280 <= contours[i][
point][0][1] <= 299:
closestContours.append(i)
break
midpointx = []
midpointy = []
mptheta = []
clcont = []
#print(len(closestContours))
if len(closestContours) == 0: # Not on line
if len(contours) >= 1:
for k in range(len(contours)): # iterate through list of contours
clcont.append(
abs(cv2.pointPolygonTest(contours[k], (200, 299), True))
) # calculate the distance between bottom centre of image and contour
followContour = clcont.index(
min(clcont)
) # find the contour with the shortest distance from point
#randpoint = [0, 0]
p = [200, 299]
#while distance.euclidean(p, randpoint) >= 150 or distance.euclidean(p, randpoint) <= 30: # choose a random point on contour
randindex = randint(0, len(contours[followContour]) - 1)
randpoint = contours[followContour][randindex][0]
for k in range(-2,
2): # choose 2 points either side of chosen point
refindex = randindex + k
if 0 <= refindex < len(
contours[followContour]
): # as long as additional points are actually on contour - if a point too close to the end is chosen, other points may fall off end of contour
midpointx.append(contours[followContour][refindex][0][0])
midpointy.append(
contours[followContour][refindex][0][1]
) # creates a set of 5 points to head towards on line
# plot midpoints
midptList = np.array(list(zip(
midpointx,
midpointy))) # create one list containing x and y midpoints
midptList = midptList.reshape(
(-1, 1, 2)) # ensure midptList is in correct format
midptList = midptList.reshape(-1, 2)
b = midptList
b = np.array([b])
b = b.astype(float) # convert to float
midptList_dist = cv2.perspectiveTransform(
b, h) # apply perspective transform to midpoints
midptList_dist = midptList_dist.reshape(
(-1, 1,
2)) # reshape output array into original contour array form
# cv2.polylines(cnt_img2, np.int32([midptList_dist]), True, (0, 255, 255), 1)
# cv2.namedWindow('midpoints')
# cv2.imshow('midpoints', cnt_img2)
for midpt1 in range(
len(midptList_dist) - 1
): # create array of angles between each midpoint of the transformed midpoints (i.e. actual angles)
mptheta.append(
np.arctan2((midptList_dist[midpt1 + 1][0][1] -
midptList_dist[midpt1][0][1]),
(midptList_dist[midpt1 + 1][0][0] -
midptList_dist[midpt1][0][0])))
mptheta.append(
mptheta[len(mptheta) - 1]
) # add last midpoint angle on at end, this makes angle array same size as midpoint array
# else:
# no contours or lines in image
# Go to Nessa's code to move robot to locate new contour
if len(
closestContours
) == 1: # Can only see one contour, choose points on this line to follow
#randpoint = [200, 150]
p = [0, 0]
#while distance.euclidean(p, randpoint) >= 550 or distance.euclidean(p, randpoint) <= 3: # choose random point on contour thats not too close or too far
randindex = randint(
0,
len(contours[closestContours[0]]) -
1) # start by choosing random index on closest contour
randpoint = contours[closestContours[0]][randindex][
0] # and then find the corresponding point and check the distance
# we choose just a few points so that if there is only one horizontal contour the robot can see, it doesnt choose points along the line
# on both sides of the robot, as we dont want it to try and go in two different directions
for k in range(-2,
2): # Again choose 2 points either side of chosen point
refindex = randindex + k
if 0 <= refindex < len(contours[closestContours[0]]):
midpointx.append(
contours[closestContours[0]][refindex][0][0]
) # robot will drive to these points located on the contours
midpointy.append(contours[closestContours[0]][refindex][0][1])
# plot midpoints
midptList = np.array(
list(zip(midpointx, midpointy)
)) # combine x and y midpoint coordinate data into one list
midptList = midptList.reshape((-1, 1, 2))
midptList = midptList.reshape(-1, 2)
b = midptList
b = np.array([b])
b = b.astype(float)
midptList_dist = cv2.perspectiveTransform(
b, h) # apply perspective transform to midpoint data
midptList_dist = midptList_dist.reshape(
(-1, 1, 2)) # reshape transformed midpoints back to original shape
# cv2.polylines(cnt_img2, np.int32([midptList_dist]), True, (0, 255, 255), 1)
# cv2.namedWindow('midpoints')
# cv2.imshow('midpoints', cnt_img2)
for midpt1 in range(len(midptList_dist) -
1): # create array of angles between each midpoint
# calculate angle between each transformed midpoint - gives real life angle
mptheta.append(
np.arctan2((midptList_dist[midpt1 + 1][0][1] -
midptList_dist[midpt1][0][1]),
(midptList_dist[midpt1 + 1][0][0] -
midptList_dist[midpt1][0][0])))
mptheta.append(mptheta[len(mptheta) - 1])
if len(closestContours
) == 2: # If two contours are close to robot, then track both
# Pick one of the two contours to follow at random
followContour = random.choice(
closestContours) # choose one of two contours to follow at random
# print(followContour)
# dist2 = []
# Find the other contour that is not being followed
m = []
contours2 = list(contours) # copy contours list
contours2.remove(
contours2[followContour]) # remove the contour we are following
for contour in range(
len(contours2
)): # iterate through contours excluding followed contour
if len(
contours2[contour]
) > 19: # for contours longer than 19, create a combined list of every other contour points
contourpoints = np.reshape(np.ravel(contours2[contour]),
(-1, 2)).tolist()
m = m + contourpoints
# followContourListPosition = closestContours.index(followContour)
# otherContour = closestContours[abs(followContourListPosition-1)]
# otherContourList = np.reshape(np.ravel(contours[otherContour]), (-1, 2))
# For each point in the followed contour, find the closest point in the other contour
tree = spatial.KDTree(m) # create tree of all contour points
for pt in range(0, len(contours[followContour]),
1): # iterate through follow contour
p = contours[followContour][pt][
0] # for each point on the follow contour
pointonLine = m[tree.query(p)
[1]] # find the closest point on the other contour
lineDistance = distance.euclidean(
p, pointonLine) # find the distance between the two lines
if 5 <= lineDistance <= 130: # if distance between lines is around tape width - i.e. lines not diverging (may need to change)
midpointx.append((p[0] + pointonLine[0]) / 2) # add midpoints
midpointy.append((p[1] + pointonLine[1]) / 2)
# plot midpoints
midptList = np.array(list(
zip(midpointx,
midpointy))) # combine midpoint coordinates into one list
midptList = midptList.reshape(
(-1, 1, 2)) # reshape this new list into the original shape
midptList = midptList.reshape(-1, 2)
b = midptList
b = np.array([b])
b = b.astype(float)
midptList_dist = cv2.perspectiveTransform(
b, h
) # find perspecitve transform of the midpoints i.e. map them to real life positions
midptList_dist = midptList_dist.reshape((-1, 1, 2))
# cv2.polylines(cnt_img2, np.int32([midptList_dist]), True, (0, 255, 255), 1)
for midpt1 in range(len(midptList_dist) -
1): # create array of angles between each midpoint
# iterate through corrected midpoint list, and calculate angle between each midpoint - as midpoints have been persp.corrected, these angles will be real life angle
mptheta.append(
np.arctan2((midptList_dist[midpt1 + 1][0][1] -
midptList_dist[midpt1][0][1]),
(midptList_dist[midpt1 + 1][0][0] -
midptList_dist[midpt1][0][0])))
mptheta.append(
mptheta[len(mptheta) - 1]
) # add on last angle to list to make angle list same length as midpoint list
# cv2.namedWindow('midpoints')
# cv2.imshow('midpoints', cnt_img2)
# cv2.imwrite('midpts.jpg', cnt_img2)
if len(closestContours) == 3: # 3 contours found in radius i.e. t junction
dist3 = []
randcont = []
# ** Old Code **
# for cont3 in range(len(closestContours)):
# dist3.append(abs(cv2.pointPolygonTest(contours[closestContours[cont3]], (200, 299), True)))
# max_dist_index = dist3.index(max(dist3))
# furthestcontour = closestContours[max_dist_index]
# randcont.append(furthestcontour) # always follow new contour (starts furthest away)
# closestContours.remove(furthestcontour)
# randcont.append(random.choice(closestContours))
# followContour = random.choice(randcont) # choose at random one of other two contours to follow
# followContourListPosition = randcont.index(followContour)
# otherContour = randcont[abs(followContourListPosition-1)] # choose at random which contour will be the main contour to follow
# otherContourList = np.reshape(np.ravel(contours[otherContour]), (-1, 2))
# **New Code **
correctcontour = 0
while correctcontour == 0:
followContour = random.choice(
closestContours) # choose random contour to follow
# print(followContour)
# Find the other contour that is not being followed
for cnt3 in range(len(contours)): # iterate through contour list
dist3 = []
if cnt3 != followContour: # don't choose the contour that is being followed
for c in range(
0, len(contours[followContour]),
round(len(contours[followContour]) /
10)): # choose 10 points along contour
fcPt = contours[followContour][c][
0] # points on follow contour
dist3.append(
abs(
cv2.pointPolygonTest(contours[cnt3],
tuple(fcPt.tolist()),
True))
) # find distance between follow contour and these 10 points
if all(
10 <= dst <= 130 for dst in dist3
): # if all distances are within a set distance, assume the contours are parallel
cnt3list = np.reshape(np.ravel(contours[cnt3]),
(-1, 2))
tree = spatial.KDTree(cnt3list)
pointonLine = cnt3list[tree.query(
fcPt)[1]] # find nearest point on other contour
midx = int(round((fcPt[0] + pointonLine[0]) / 2))
midy = int(round((fcPt[1] + pointonLine[1]) / 2))
midpt = [midx, midy]
if resized[midy][
midx] < 100: # check that this midpoint is on a black line, and not the white gap between two lines
otherContour = cnt3
correctcontour = 1 # set flag to break from while loop
break
otherContourList = np.reshape(np.ravel(contours[otherContour]),
(-1, 2))
# For each point in the followed contour, find the closest point in the other contour
tree = spatial.KDTree(
otherContourList) # create tree of points on other contour
for pt in range(len(
contours[followContour])): # iterate through contour followed
p = contours[followContour][pt][0]
pointonLine = otherContourList[tree.query(
p
)[1]] # find closest point on other contour to each point on follow contour
lineDistance = distance.euclidean(p, pointonLine)
if 5 <= lineDistance <= 150: # if distance between lines is around tape width - i.e. lines not diverging (may need to change)
midpointx.append((p[0] + pointonLine[0]) / 2) # add midpoints
midpointy.append((p[1] + pointonLine[1]) / 2)
# plot midpoints
midptList = np.array(list(zip(
midpointx, midpointy))) # join x and y midpoints into one list
midptList = midptList.reshape(-1, 2)
b = midptList
b = np.array([b])
b = b.astype(float)
midptList_dist = cv2.perspectiveTransform(
b, h) # apply perspective transforms to midpoints
midptList_dist = midptList_dist.reshape((-1, 1, 2))
#cv2.polylines(cnt_img2, np.int32([midptList_dist]), True, (0, 255, 255), 1)
for midpt1 in range(len(midptList_dist) -
1): # create array of angles between each midpoint
# find angle between each midpoint
mptheta.append(
np.arctan2((midptList_dist[midpt1 + 1][0][1] -
midptList_dist[midpt1][0][1]),
(midptList_dist[midpt1 + 1][0][0] -
midptList_dist[midpt1][0][0])))
mptheta.append(mptheta[len(mptheta) - 1])
#cv2.namedWindow('midpoints')
# cv2.imshow('midpoints', cnt_img2)
#cv2.imwrite('midpointsnew',)
if len(closestContours) > 3:
for k in range(len(contours)): # iterate through contours
clcont.append(
abs(cv2.pointPolygonTest(contours[k], (200, 299), True))
) # create array of distances between robot and each contour
followContour = clcont.index(
min(clcont)) # choose the closest contour to follow
#randpoint = [0, 0]
p = [200, 299]
#while distance.euclidean(p, randpoint) >= 250 or distance.euclidean(p, randpoint) <= 30:
randindex = randint(0, len(contours[followContour]) - 1)
randpoint = contours[followContour][randindex][0]
for k in range(
-2,
2): # choose some points on the closest contour to follow
refindex = randindex + k
if 0 <= refindex < len(contours[followContour]):
midpointx.append(contours[followContour][refindex][0][0])
midpointy.append(contours[followContour][refindex][0][1])
# plot midpoints
midptList = np.array(list(zip(midpointx, midpointy)))
midptList = midptList.reshape((-1, 1, 2))
midptList = midptList.reshape(-1, 2)
b = midptList
b = np.array([b])
b = b.astype(float)
midptList_dist = cv2.perspectiveTransform(b, h)
midptList_dist = midptList_dist.reshape((-1, 1, 2))
# cv2.polylines(cnt_img2, np.int32([midptList_dist]), True, (0, 255, 255), 1)
# cv2.namedWindow('midpoints')
# cv2.imshow('midpoints', cnt_img2)
for midpt1 in range(
len(midptList_dist) -
1): # create array of angles between each midpoint
mptheta.append(
np.arctan2((midptList_dist[midpt1 + 1][0][1] -
midptList_dist[midpt1][0][1]),
(midptList_dist[midpt1 + 1][0][0] -
midptList_dist[midpt1][0][0])))
mptheta.append(mptheta[len(mptheta) - 1])
# end_time = time.time()-start_time
#print('finished')
return [midptList_dist,
mptheta] # return list of x and y midpoints and angle list
##test:
#img = cv2.imread('test5.jpg',0)
#[midptList_dist, mptheta] = imgProcess(img)
def main(self):
#model
prototxt = 'MobileNetSSD_deploy.prototxt.txt'
model = 'MobileNetSSD_deploy.caffemodel'
CLASSES = ["car"]
# load our serialized model from disk
print("[INFO] loading model...")
net = cv2.dnn.readNetFromCaffe(prototxt, model)
# m = coordinates
m = ((48, 443), (196, 367), (446, 581), (243, 685))
# Read a image
I = cv2.imread('images/image.jpg')
# I = frame
# First find the minX minY maxX and maxY of the polygon
minX = I.shape[1]
maxX = -1
minY = I.shape[0]
maxY = -1
for point in m:
x = point[0]
y = point[1]
if x < minX:
minX = x
if x > maxX:
maxX = x
if y < minY:
minY = y
if y > maxY:
maxY = y
# Go over the points in the image if thay are out side of the emclosing rectangle put zero
# if not check if thay are inside the polygon or not
cropedImage = np.zeros_like(I)
for y in range(0, I.shape[0]):
for x in range(0, I.shape[1]):
if x < minX or x > maxX or y < minY or y > maxY:
continue
if cv2.pointPolygonTest(np.asarray(m), (x, y), False) >= 0:
cropedImage[y, x, 0] = I[y, x, 0]
cropedImage[y, x, 1] = I[y, x, 1]
cropedImage[y, x, 2] = I[y, x, 2]
# Now we can crop again just the envloping rectangle
finalImage = cropedImage[minY:maxY, minX:maxX]
image = finalImage
(h, w) = image.shape[:2]
blob = cv2.dnn.blobFromImage(cv2.resize(image, (300, 300)), 0.007843,
(300, 300), 127.5)
# pass the blob through the network and obtain the detections and
# predictions
print("[INFO] computing object detections...")
net.setInput(blob)
detections = net.forward()
# loop over the detections
for i in np.arange(0, detections.shape[2]):
confidence = detections[0, 0, i, 2]
# print type(confidence)
if confidence >= 0.2:
print 'confidence'
else:
print 'no detection'
cv2.imshow('d', finalImage)
def acute(obj, mask, win, thresh):
"""acute: identify landmark positions within a contour for morphometric analysis
Inputs:
obj = An opencv contour array of interest to be scanned for landmarks
mask = binary mask used to generate contour array (necessary for ptvals)
win = maximum cumulative pixel distance window for calculating angle
score; 1 cm in pixels often works well
thresh = angle score threshold to be applied for mapping out landmark
coordinate clusters within each contour
Outputs:
homolog_pts = pseudo-landmarks selected from each landmark cluster
start_pts = pseudo-landmark island starting position; useful in parsing homolog_pts in downstream analyses
stop_pts = pseudo-landmark island end position ; useful in parsing homolog_pts in downstream analyses
ptvals = average values of pixel intensity from the mask used to generate cont;
useful in parsing homolog_pts in downstream analyses
chain = raw angle scores for entire contour, used to visualize landmark
clusters
verbose_out = supplemental file which stores coordinates, distance from
landmark cluster edges, and angle score for entire contour. Used
in troubleshooting.
:param obj: ndarray
:param mask: ndarray
:param win: int
:param thresh: int
:return homolog_pts:
"""
chain = [] # Create empty chain to store angle scores
for k in list(range(len(obj))): # Coordinate-by-coordinate 3-point assignments
vert = obj[k]
dist_1 = 0
for r in range(len(obj)): # Reverse can to obtain point A
rev = k - r
pos = obj[rev]
dist_2 = np.sqrt(np.square(pos[0][0]-vert[0][0])+np.square(pos[0][1]-vert[0][1]))
if r >= 2:
if (dist_2 > dist_1) & (dist_2 <= win): # Further from vertex than current pt A while within window?
dist_1 = dist_2
ptA = pos # Load best fit within window as point A
elif dist_2 > win:
break
else:
ptA = pos
dist_1 = 0
for f in range(len(obj)): # Forward scan to obtain point B
fwd = k + f
if fwd >= len(obj):
fwd -= len(obj)
pos = obj[fwd]
dist_2 = np.sqrt(np.square(pos[0][0]-vert[0][0])+np.square(pos[0][1]-vert[0][1]))
if f >= 2:
if (dist_2 > dist_1) & (dist_2 <= win): # Further from vertex than current pt B while within window?
dist_1 = dist_2
ptB = pos # Load best fit within window as point B
elif dist_2 > win:
break
else:
ptB = pos
# Angle in radians derived from Law of Cosines, converted to degrees
P12 = np.sqrt((vert[0][0]-ptA[0][0])*(vert[0][0]-ptA[0][0])+(vert[0][1]-ptA[0][1])*(vert[0][1]-ptA[0][1]))
P13 = np.sqrt((vert[0][0]-ptB[0][0])*(vert[0][0]-ptB[0][0])+(vert[0][1]-ptB[0][1])*(vert[0][1]-ptB[0][1]))
P23 = np.sqrt((ptA[0][0]-ptB[0][0])*(ptA[0][0]-ptB[0][0])+(ptA[0][1]-ptB[0][1])*(ptA[0][1]-ptB[0][1]))
dot = (P12*P12 + P13*P13 - P23*P23)/(2*P12*P13)
# Used a random number generator to test if either of these cases were possible but neither is possible
# if dot > 1: # If float exceeds 1 prevent arcos error and force to equal 1
# dot = 1
# elif dot < -1: # If float exceeds -1 prevent arcos error and force to equal -1
# dot = -1
ang = math.degrees(math.acos(dot))
chain.append(ang)
index = [] # Index chain to find clusters below angle threshold
for c in range(len(chain)): # Identify links in chain with acute angles
if float(chain[c]) <= thresh:
index.append(c) # Append positions of acute links to index
acute_pos = obj[index] # Extract all island points blindly
float(len(acute_pos)) / float(len(obj)) # Proportion of informative positions
if len(index) != 0:
isle = []
island = []
for c in range(len(index)): # Scan for iterative links within index
if not island:
island.append(index[c]) # Initiate new link island
elif island[-1]+1 == index[c]:
island.append(index[c]) # Append successful iteration to island
elif island[-1]+1 != index[c]:
ptA = obj[index[c]]
ptB = obj[island[-1]+1]
dist = np.sqrt(np.square(ptA[0][0]-ptB[0][0])+np.square(ptA[0][1]-ptB[0][1]))
if win/2 > dist:
island.append(index[c])
else:
isle.append(island)
island = [index[c]]
isle.append(island)
if len(isle) > 1:
if (isle[0][0] == 0) & (isle[-1][-1] == (len(chain)-1)):
print('Fusing contour edges')
# Cannot add a range and a list (or int)
# island = range(-(len(chain)-isle[-1][0]), 0)+isle[0] # Fuse overlapping ends of contour
# Delete islands to be spliced if start-end fusion required
del isle[0]
del isle[-1]
# isle.insert(0, island) # Prepend island to isle
else:
print('Microcontour...')
# Homologous point maximum distance method
pt = []
vals = []
maxpts = []
SSpts = []
TSpts = []
ptvals = []
max_dist = [['cont_pos', 'max_dist', 'angle']]
for x in range(len(isle)):
# Identify if contour is concavity/convexity using image mask
pix_x, pix_y, w, h = cv2.boundingRect(obj[isle[x]]) # Obtain local window around island
for c in range(w):
for r in range(h):
# Identify pixels in local window internal to the island hull
pos = cv2.pointPolygonTest(obj[isle[x]], (pix_x+c, pix_y+r), 0)
if 0 < pos:
vals.append(mask[pix_y+r][pix_x+c]) # Store pixel value if internal
if len(vals) > 0:
ptvals.append(sum(vals)/len(vals))
vals = []
else:
ptvals.append('NaN') # If no values can be retrieved (small/collapsed contours)
vals = []
# Identify pixel coordinate to use as pseudolandmark for island
if len(isle[x]) == 1: # If landmark is a single point (store position)
# print 'route A'
pt = isle[x][0]
max_dist.append([isle[x][0], '-', chain[isle[x][0]]])
# print pt
elif len(isle[x]) == 2: # If landmark is a pair of points (store more acute position)
# print 'route B'
ptA = chain[isle[x][0]]
ptB = chain[isle[x][1]]
if ptA < ptB:
pt = isle[x][0] # Store point A if more acute
max_dist.append([isle[x][0], '-', chain[isle[x][0]]])
elif ptA > ptB:
pt = isle[x][1] # Store point B if more acute
max_dist.append([isle[x][1], '-', chain[isle[x][1]]])
# print pt
else: # If landmark is multiple points (distance scan for position)
# print 'route C'
SS = obj[[isle[x]]][0] # Store isle "x" start site
TS = obj[[isle[x]]][-1] # Store isle "x" termination site
dist_1 = 0
for d in range(len(isle[x])): # Scan from SS to TS within isle "x"
site = obj[[isle[x][d]]]
SSd = np.sqrt(np.square(SS[0][0]-site[0][0][0])+np.square(SS[0][1]-site[0][0][1]))
TSd = np.sqrt(np.square(TS[0][0]-site[0][0][0])+np.square(TS[0][1]-site[0][0][1]))
# Current mean distance of 'd' to 'SS' & 'TS'
dist_2 = np.mean([np.abs(SSd), np.abs(TSd)])
max_dist.append([isle[x][d], dist_2, chain[isle[x][d]]])
if dist_2 > dist_1: # Current mean distance better fit that previous best?
pt = isle[x][d]
dist_1 = dist_2 # Current mean becomes new best mean
# print pt
maxpts.append(pt) # Empty 'pts' prior to next mean distance scan
SSpts.append(isle[x][0])
TSpts.append(isle[x][-1])
homolog_pts = obj[maxpts]
start_pts = obj[SSpts]
stop_pts = obj[TSpts]
return homolog_pts, start_pts, stop_pts, ptvals, chain, max_dist
else:
return [], [], [], [], [], []
plt.imshow(NewMaxFlare)
#contours on MinFlare image
_, contours2, _ = cv2.findContours(MinFlare.astype('uint8'), cv2.RETR_TREE,
cv2.CHAIN_APPROX_SIMPLE)
NewMinFlare = cv2.drawContours(MinFlare.astype('uint8'), contours2, -1,
(0, 0, 255), 1)
plt.figure()
plt.title(' contours on AIA 1700 (at time of flare minimum) ')
plt.imshow(NewMinFlare)
area = []
for rows in range(MaxFlare.shape[0]):
for columns in range(MaxFlare.shape[1]):
for n in contours:
if (cv2.pointPolygonTest(n, (rows, columns),
measureDist=False)) == 1:
area.append([rows, columns])
#contours on Magnetogram (max)
NewMag = cv2.drawContours(Mag.astype('uint8'), contours, -1, (0, 0, 255), 1)
plt.figure()
plt.title(' contours on Magnetogram (at time of flare maximum) ')
plt.imshow(NewMag)
#contours on Continuum
NewCon = cv2.drawContours(Con.astype('uint8'), contours, -1, (0, 0, 255), 1)
plt.figure()
plt.title(' contours on Continuum (at time of flare maximum) ')
plt.imshow(NewCon)
#contours on Magnetogram (min)
NewMag2 = cv2.drawContours(Mag.astype('uint8'), contours2, -1, (0, 0, 255), 1)
def __iter__(self):
for x in range(self.min[0], self.max[0] + 1):
for y in range(self.min[1], self.max[1] + 1):
if cv2.pointPolygonTest(self._contour, (x, y),
measureDist=False) >= 0:
yield x, y
def random_warp_landmarks(self, image, src_points=None, dst_points=None):
""" get warped image, target image and target mask
From DFAKER plugin """
logger.trace("Randomly warping landmarks")
size = image.shape[0]
coverage = self.get_coverage(image) // 2
p_mx = size - 1
p_hf = (size // 2) - 1
edge_anchors = [(0, 0), (0, p_mx), (p_mx, p_mx), (p_mx, 0), (p_hf, 0),
(p_hf, p_mx), (p_mx, p_hf), (0, p_hf)]
grid_x, grid_y = np.mgrid[0:p_mx:complex(size), 0:p_mx:complex(size)]
source = src_points
destination = (dst_points.copy().astype('float32') +
np.random.normal(size=dst_points.shape, scale=2.0))
destination = destination.astype('uint8')
face_core = cv2.convexHull(
np.concatenate( # pylint:disable=no-member
[source[17:], destination[17:]],
axis=0).astype(int))
source = [(pty, ptx) for ptx, pty in source] + edge_anchors
destination = [(pty, ptx) for ptx, pty in destination] + edge_anchors
indicies_to_remove = set()
for fpl in source, destination:
for idx, (pty, ptx) in enumerate(fpl):
if idx > 17:
break
elif cv2.pointPolygonTest(
face_core, # pylint:disable=no-member
(pty, ptx),
False) >= 0:
indicies_to_remove.add(idx)
for idx in sorted(indicies_to_remove, reverse=True):
source.pop(idx)
destination.pop(idx)
grid_z = griddata(destination,
source, (grid_x, grid_y),
method="linear")
map_x = np.append([], [ar[:, 1] for ar in grid_z]).reshape(size, size)
map_y = np.append([], [ar[:, 0] for ar in grid_z]).reshape(size, size)
map_x_32 = map_x.astype('float32')
map_y_32 = map_y.astype('float32')
warped_image = cv2.remap(
image, # pylint:disable=no-member
map_x_32,
map_y_32,
cv2.INTER_LINEAR, # pylint:disable=no-member
cv2.BORDER_TRANSPARENT) # pylint:disable=no-member
target_image = image
# TODO Make sure this replacement is correct
slices = slice(size // 2 - coverage, size // 2 + coverage)
# slices = slice(size // 32, size - size // 32) # 8px on a 256px image
warped_image = cv2.resize( # pylint:disable=no-member
warped_image[slices, slices, :],
(self.input_size, self.input_size), cv2.INTER_AREA) # pylint:disable=no-member
logger.trace("Warped image shape: %s", warped_image.shape)
target_images = [
cv2.resize(
target_image[slices, slices, :], # pylint:disable=no-member
(size, size),
cv2.INTER_AREA) # pylint:disable=no-member
for size in self.output_sizes
]
logger.trace("Target image shapea: %s",
[img.shape for img in target_images])
return self.compile_images(warped_image, target_images)
def create_cloth_mesh(cloth_img, cloth_contours):
points = []
height, width, _ = cloth_img.shape
# 1.1 get the biggest/external contour
maxcontouridx = 0
maxcontourlen = 0
for i in range(len(cloth_contours)):
if maxcontourlen < len(cloth_contours[i]):
maxcontourlen = len(cloth_contours[i])
maxcontouridx = i
max_contour = cloth_contours[maxcontouridx]
# get all mesh points/vertices and handle points
# note control vertices
handle_v_list = []
# 1.2 add sampled points from contour
# seglen = maxcontourlen//20
seglen = maxcontourlen // 30
vidx = 0
for ind, each in enumerate(max_contour):
if ind % seglen == 0 and ind > 0 and not _checkCloseToHandles(
handle_v_list, points, each[0]):
if vidx not in handle_v_list:
handle_v_list.append(
vidx) # now we add only contours for handles
points.append(tuple(each[0])) # add mesh vertices also
vidx = vidx + 1
else:
try:
# check angles of the points, take acute or smaller obtuse angles for adding to control points
this_angle = get_angle(max_contour[ind - 5][0],
max_contour[ind][0],
max_contour[ind + 5][0])
if this_angle < 150 and not _checkCloseToHandles(
handle_v_list, points, each[0]):
# if this_angle < 150:
if vidx not in handle_v_list:
handle_v_list.append(
vidx) # now we add only contours for handles
points.append(tuple(each[0])) # add mesh vertices also
vidx = vidx + 1
except Exception as err:
print(err)
# 2. the bounding box
# get minimum and maximum of the contour
mc_x = max_contour[:, :, 0]
mc_y = max_contour[:, :, 1]
xmin = min(mc_x)[0]
ymin = min(mc_y)[0]
xmax = max(mc_x)[0]
ymax = max(mc_y)[0]
seglen = (xmax - xmin) // 10 # 10 # 20
# add points from inside cloth
for _y in range(int((ymax - ymin) / seglen)):
for _x in range(int((xmax - xmin) / seglen)):
x, y = xmin + seglen * _x, ymin + seglen * _y # bug fixed 2020. 8. 16
if ymin <= y <= ymax and xmin <= x <= xmax:
# >= 0: # check if inside cloth contour
dist = cv2.pointPolygonTest(max_contour, (x, y), True)
if dist > 0 and dist > seglen / 4: # inside and not too close to the contours
points.append((x, y))
# 3 list to numpy array @TODO
o_vertices = np.asarray(points)
o_handles = np.asarray(handle_v_list)
# 4 build triangles
# @Note: now we generate rectangle mesh only, so do not need to use Subdiv2D
rect = (xmin, ymin, xmax, ymax)
# @TODO Do we need to use opencv ? We could build more easily
subdiv = cv2.Subdiv2D(rect)
for p in points:
subdiv.insert(p)
# why we get points for minus coordinates
triangleList = subdiv.getTriangleList()
# 5 build triangles (of indices of vertices) from point locations
triangles = np.zeros((len(triangleList), 3), dtype=int)
tidx = 0
for t in triangleList:
x, y = t[0], t[1]
if (xmin > x or x > xmax) or (
ymin > y or y > ymax): # often subdiv2d gives out of rectangle
continue
x, y = t[2], t[3]
if (xmin > x or x > xmax) or (
ymin > y or y > ymax): # often subdiv2d gives out of rectangle
continue
x, y = t[4], t[5]
if (xmin > x or x > xmax) or (
ymin > y or y > ymax): # often subdiv2d gives out of rectangle
continue
idx0 = _findNearestinMesh(o_vertices, (t[0], t[1]))
idx1 = _findNearestinMesh(o_vertices, (t[2], t[3]))
idx2 = _findNearestinMesh(o_vertices, (t[4], t[5]))
# get the triangle center
if False:
centerX = (points[idx0][0] + points[idx1][0] + points[idx2][0]) / 3
centerY = (points[idx0][1] + points[idx1][1] + points[idx2][1]) / 3
# check if inside cloth contour
if cv2.pointPolygonTest(max_contour,
(centerX, centerY), True) >= 0:
triangles[tidx] = (idx0, idx1, idx2)
tidx = tidx + 1
else:
triangles[tidx] = (idx0, idx1, idx2)
tidx = tidx + 1
triangles = np.resize(triangles, (tidx, 3)) # remove triangle out of cloth
# 3. Finally create meshes and handle points objects
o_mesh = TriangleMesh(o_vertices, triangles, o_handles)
handle_tracker = ControlPtsTrack(handle_v_list)
for each in handle_v_list:
handle_tracker.srcPos.append(o_mesh.vertices[each, :])
handle_tracker.tgtPos = handle_tracker.srcPos.copy()
return o_mesh, handle_tracker
def isOutside(self, point):
dist = cv2.pointPolygonTest(self.contour, point, True)
if dist < 0:
return True
else:
return False
def __call__(self, pt):
return 1 if cv2.pointPolygonTest(self.cont, pt, False) >= 0 else 0
def analyze_bound_vertical(img, obj, mask, line_position):
"""User-input boundary line tool
Inputs:
img = RGB or grayscale image data for plotting
obj = single or grouped contour object
mask = Binary mask made from selected contours
shape_header = pass shape header data to function
shape_data = pass shape data so that analyze_bound data can be appended to it
line_position = position of boundary line (a value of 0 would draw the line through the left side of the image)
Returns:
analysis_images = output images
:param img: numpy.ndarray
:param obj: list
:param mask: numpy.ndarray
:param line_position: int
:return analysis_images: list
"""
ori_img = np.copy(img)
# Draw line horizontal line through bottom of image, that is adjusted to user input height
if len(np.shape(ori_img)) == 2:
ori_img = cv2.cvtColor(ori_img, cv2.COLOR_GRAY2BGR)
iy, ix, iz = np.shape(ori_img)
size = (iy, ix)
size1 = (iy, ix, 3)
background = np.zeros(size, dtype=np.uint8)
wback = (np.zeros(size1, dtype=np.uint8)) + 255
x_coor = 0 + int(line_position)
y_coor = int(iy)
rec_point1 = (0, 0)
rec_point2 = (x_coor, y_coor - 2)
cv2.rectangle(background, rec_point1, rec_point2, (255), -1)
right_contour, right_hierarchy = cv2.findContours(
background, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
x, y, width, height = cv2.boundingRect(obj)
if x_coor - x <= 0:
width_left_bound = 0
width_right_bound = width
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
width_left_bound = width
width_right_bound = 0
else:
width_left_bound = x_coor - x
width_right_bound = width - width_left_bound
right = []
left = []
mask_nonzerox, mask_nonzeroy = np.nonzero(mask)
obj_points = np.vstack((mask_nonzeroy, mask_nonzerox))
obj_points1 = np.transpose(obj_points)
for i, c in enumerate(obj_points1):
xy = tuple(c)
pptest = cv2.pointPolygonTest(right_contour[0], xy, measureDist=False)
if pptest == 1:
left.append(xy)
cv2.circle(ori_img, xy, 1, (155, 0, 255))
cv2.circle(wback, xy, 1, (155, 0, 255))
else:
right.append(xy)
cv2.circle(ori_img, xy, 1, (0, 255, 0))
cv2.circle(wback, xy, 1, (0, 255, 0))
right_bound_area = len(right)
left_bound_area = len(left)
percent_bound_area_right = (
(float(right_bound_area)) /
(float(left_bound_area + right_bound_area))) * 100
percent_bound_area_left = (
(float(left_bound_area)) /
(float(right_bound_area + left_bound_area))) * 100
analysis_images = []
if left_bound_area or right_bound_area:
point3 = (x_coor + 2, 0)
point4 = (x_coor + 2, y_coor)
cv2.line(ori_img, point3, point4, (255, 0, 255), params.line_thickness)
cv2.line(wback, point3, point4, (255, 0, 255), params.line_thickness)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if x_coor - x <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)),
(0, 255, 0), params.line_thickness)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0),
params.line_thickness)
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)),
(255, 0, 0), params.line_thickness)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)),
(255, 0, 0), params.line_thickness)
else:
cv2.line(ori_img, (x_coor + 2, int(cmy)),
(x_coor + width_left_bound, int(cmy)), (255, 0, 0),
params.line_thickness)
cv2.line(ori_img, (x_coor + 2, int(cmy)),
(x_coor - width_right_bound, int(cmy)), (0, 255, 0),
params.line_thickness)
cv2.line(wback, (x_coor + 2, int(cmy)),
(x_coor + width_left_bound, int(cmy)), (255, 0, 0),
params.line_thickness)
cv2.line(wback, (x_coor + 2, int(cmy)),
(x_coor - width_right_bound, int(cmy)), (0, 255, 0),
params.line_thickness)
# Output images with boundary line
analysis_images.append(wback)
analysis_images.append(ori_img)
if params.debug is not None:
params.device += 1
point3 = (x_coor + 2, 0)
point4 = (x_coor + 2, y_coor)
cv2.line(ori_img, point3, point4, (255, 0, 255), params.line_thickness)
cv2.line(wback, point3, point4, (255, 0, 255), params.line_thickness)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if x_coor - x <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)),
(0, 255, 0), params.line_thickness)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)), (0, 255, 0),
params.line_thickness)
elif x_coor - x > 0:
width_1 = x_coor - x
if width - width_1 <= 0:
cv2.line(ori_img, (x, int(cmy)), (x + width, int(cmy)),
(255, 0, 0), params.line_thickness)
cv2.line(wback, (x, int(cmy)), (x + width, int(cmy)),
(255, 0, 0), params.line_thickness)
else:
cv2.line(ori_img, (x_coor + 2, int(cmy)),
(x_coor + width_left_bound, int(cmy)), (255, 0, 0),
params.line_thickness)
cv2.line(ori_img, (x_coor + 2, int(cmy)),
(x_coor - width_right_bound, int(cmy)), (0, 255, 0),
params.line_thickness)
cv2.line(wback, (x_coor + 2, int(cmy)),
(x_coor + width_left_bound, int(cmy)), (255, 0, 0),
params.line_thickness)
cv2.line(wback, (x_coor + 2, int(cmy)),
(x_coor - width_right_bound, int(cmy)), (0, 255, 0),
params.line_thickness)
if params.debug == 'print':
print_image(
wback,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_white.png'))
print_image(
ori_img,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_img.png'))
if params.debug == 'plot':
plot_image(wback)
plot_image(ori_img)
outputs.add_observation(variable='vertical_reference_position',
trait='vertical reference position',
method='plantcv.plantcv.analyze_bound_vertical',
scale='none',
datatype=int,
value=line_position,
label='none')
outputs.add_observation(variable='width_left_reference',
trait='width left of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='pixels',
datatype=int,
value=width_left_bound,
label='pixels')
outputs.add_observation(variable='width_right_reference',
trait='width right of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='pixels',
datatype=int,
value=width_right_bound,
label='pixels')
outputs.add_observation(variable='area_left_reference',
trait='area left of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='pixels',
datatype=int,
value=left_bound_area,
label='pixels')
outputs.add_observation(variable='percent_area_left_reference',
trait='percent area left of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='none',
datatype=float,
value=percent_bound_area_left,
label='none')
outputs.add_observation(variable='area_right_reference',
trait='area right of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='pixels',
datatype=int,
value=right_bound_area,
label='pixels')
outputs.add_observation(variable='percent_area_right_reference',
trait='percent area right of reference',
method='plantcv.plantcv.analyze_bound_vertical',
scale='none',
datatype=float,
value=percent_bound_area_right,
label='none')
# Store images
outputs.images.append(analysis_images)
return ori_img
def analyze_bound_horizontal(img, obj, mask, line_position, filename=False):
"""User-input boundary line tool
Inputs:
img = RGB or grayscale image data for plotting
obj = single or grouped contour object
mask = Binary mask made from selected contours
line_position = position of boundry line (a value of 0 would draw the line through the bottom of the image)
filename = False or image name. If defined print image.
Returns:
bound_header = data table column headers
bound_data = boundary data table
analysis_images = output image filenames
:param img: numpy.ndarray
:param obj: list
:param mask: numpy.ndarray
:param line_position: int
:param filename: str
:return bound_header: tuple
:return bound_data: tuple
:return analysis_images: list
"""
params.device += 1
ori_img = np.copy(img)
# Draw line horizontal line through bottom of image, that is adjusted to user input height
if len(np.shape(ori_img)) == 2:
ori_img = cv2.cvtColor(ori_img, cv2.COLOR_GRAY2BGR)
iy, ix, iz = np.shape(ori_img)
size = (iy, ix)
size1 = (iy, ix, 3)
background = np.zeros(size, dtype=np.uint8)
wback = (np.zeros(size1, dtype=np.uint8)) + 255
x_coor = int(ix)
y_coor = int(iy) - int(line_position)
rec_corner = int(iy - 2)
rec_point1 = (1, rec_corner)
rec_point2 = (x_coor - 2, y_coor - 2)
cv2.rectangle(background, rec_point1, rec_point2, (255), 1)
below_contour, below_hierarchy = cv2.findContours(
background, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
x, y, width, height = cv2.boundingRect(obj)
if y_coor - y <= 0:
height_above_bound = 0
height_below_bound = height
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
height_above_bound = height
height_below_bound = 0
else:
height_above_bound = y_coor - y
height_below_bound = height - height_above_bound
below = []
above = []
mask_nonzerox, mask_nonzeroy = np.nonzero(mask)
obj_points = np.vstack((mask_nonzeroy, mask_nonzerox))
obj_points1 = np.transpose(obj_points)
for i, c in enumerate(obj_points1):
xy = tuple(c)
pptest = cv2.pointPolygonTest(below_contour[0], xy, measureDist=False)
if pptest == 1:
below.append(xy)
cv2.circle(ori_img, xy, 1, (0, 0, 255))
cv2.circle(wback, xy, 1, (0, 0, 255))
else:
above.append(xy)
cv2.circle(ori_img, xy, 1, (0, 255, 0))
cv2.circle(wback, xy, 1, (0, 255, 0))
above_bound_area = len(above)
below_bound_area = len(below)
percent_bound_area_above = (
(float(above_bound_area)) /
(float(above_bound_area + below_bound_area))) * 100
percent_bound_area_below = (
(float(below_bound_area)) /
(float(above_bound_area + below_bound_area))) * 100
bound_header = [
'HEADER_BOUNDARY' + str(line_position), 'height_above_bound',
'height_below_bound', 'above_bound_area', 'percent_above_bound_area',
'below_bound_area', 'percent_below_bound_area'
]
bound_data = [
'BOUNDARY_DATA', height_above_bound, height_below_bound,
above_bound_area, percent_bound_area_above, below_bound_area,
percent_bound_area_below
]
analysis_images = []
if above_bound_area or below_bound_area:
point3 = (0, y_coor - 4)
point4 = (x_coor, y_coor - 4)
cv2.line(ori_img, point3, point4, (255, 0, 255), 5)
cv2.line(wback, point3, point4, (255, 0, 255), 5)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if y_coor - y <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(0, 255, 0), 3)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height), (0, 255, 0),
3)
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), 3)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), 3)
else:
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
3)
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
3)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
3)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
3)
if filename:
# Output images with boundary line, above/below bound area
out_file = str(
filename[0:-4]) + '_boundary' + str(line_position) + '.jpg'
print_image(ori_img, out_file)
analysis_images = ['IMAGE', 'boundary', out_file]
if params.debug is not None:
point3 = (0, y_coor - 4)
point4 = (x_coor, y_coor - 4)
cv2.line(ori_img, point3, point4, (255, 0, 255), 5)
cv2.line(wback, point3, point4, (255, 0, 255), 5)
m = cv2.moments(mask, binaryImage=True)
cmx, cmy = (m['m10'] / m['m00'], m['m01'] / m['m00'])
if y_coor - y <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(0, 255, 0), 3)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height), (0, 255, 0),
3)
elif y_coor - y > 0:
height_1 = y_coor - y
if height - height_1 <= 0:
cv2.line(ori_img, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), 3)
cv2.line(wback, (int(cmx), y), (int(cmx), y + height),
(255, 0, 0), 3)
else:
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
3)
cv2.line(ori_img, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
3)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor - height_above_bound), (255, 0, 0),
3)
cv2.line(wback, (int(cmx), y_coor - 2),
(int(cmx), y_coor + height_below_bound), (0, 255, 0),
3)
if params.debug == 'print':
print_image(
wback,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_white.jpg'))
print_image(
ori_img,
os.path.join(params.debug_outdir,
str(params.device) + '_boundary_on_img.jpg'))
if params.debug == 'plot':
plot_image(wback)
plot_image(ori_img)
return bound_header, bound_data, analysis_images
def step1():
cell_area_hist_list = []
print("============Step 1 Start============")
#-----read-----
root = tk.Tk()
root.withdraw()
file_path = filedialog.askopenfilename()
#img=cv.imread("G:\\2020summer\\Project\\Chromophobe_dataset1\\4.jpg")
img = cv.imread(file_path)
img_original = img
print("Img size: [Height :", img.shape[0], "]", "[Width :", img.shape[1],
"]")
if img.shape[0] < 450 or img.shape[1] < 600:
img = cv.copyMakeBorder(img,
80,
450,
360,
360,
cv.BORDER_CONSTANT,
value=[255, 255, 255])
else:
img = cv.copyMakeBorder(img,
80,
450,
60,
60,
cv.BORDER_CONSTANT,
value=[255, 255, 255])
#img=cv.cvtColor(img,cv.COLOR_BGR2BGRA)
img_masked = img.copy()
img_nucleus_white_img = img.copy()
#-----preprocess-----
gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
#cv.imshow("gray", gray)
gauss = cv.GaussianBlur(gray, (5, 5), 5)
#cv.imshow("gauss1",gauss)
ret, thresh = cv.threshold(gauss, 190, 255, 0)
cv.imwrite(
"G:\\2020summer\\Project\\Chromophobe_dataset1\\figure3_left.jpg",
thresh)
#cv.imshow("thresh",thresh)
erode = cv.erode(thresh, None, iterations=1)
#cv.imshow("erode",erode)
#-----remove outlines-----
#cv.imshow("erode",erode)
for i in range(0, img.shape[0]):
for j in range(0, img.shape[1]):
erode[0][j] = 255
#-----find contours-----
cnts, hierarchy = cv.findContours(erode.copy(), cv.RETR_LIST,
cv.CHAIN_APPROX_NONE)
def cnt_area(cnt):
area = cv.contourArea(cnt)
return area
counter_number = 0
location_cells_center = {}
area_of_cells_nucleus = []
Whole_pic_cell_area_ave_percent = []
Whole_pic_cell_color_ave = []
for i in range(0, len(cnts)):
if 250 <= cnt_area(cnts[i]) <= 0.2 * (img.shape[0] * img.shape[1]):
cell_area_hist_list.append(cnt_area(cnts[i]))
#print(cnts[i])
#cell_area_hist_list.append(area_calculate_from_points(cnts[i]))
counter_number += 1
#print(cnts[i])
#print("======")
cv.drawContours(img_masked, cnts[i], -1, (0, 0, 255),
2) #draw contours
cv.drawContours(img_nucleus_white_img, [cnts[i]], -1,
(255, 255, 255), -1) #masked white
M = cv.moments(cnts[i])
#检索每个细胞的内部颜色
x, y, w, h = cv.boundingRect(cnts[i])
#cv.imshow('single_cell', newimage)
cell_area_percent = 0
for row in range(y, y + h):
for col in range(x, x + w):
result = cv.pointPolygonTest(cnts[i], (col, row), False)
if result == -1:
cv.circle(gray, (col, row), 1, 255, -1)
cv.circle(img, (col, row), 1, (255, 255, 255), -1)
cell_area_percent += 1
cv.rectangle(img, (x, y), (x + w, y + h), (153, 153, 0), 1)
newimage_gray = gray[y:y + h, x:x + w]
Single_Cell_Color_Distrution = []
for row in range(h):
for col in range(w):
if newimage_gray[row, col] != 255:
Single_Cell_Color_Distrution.append(newimage_gray[row,
col])
'''
plt.hist(Single_Cell_Color_Distrution,bins=50)
plt.title(str(counter_number))
plt.show()
'''
#print("this cell area percent= ",str(cell_area_percent/(w*h)))
numpy.set_printoptions(precision=3)
Whole_pic_cell_area_ave_percent.append(cell_area_percent / (w * h))
Whole_pic_cell_color_ave.append(
numpy.mean(Single_Cell_Color_Distrution))
"""#找出masked细胞内点的坐标
rect = cv.minAreaRect(cnts[i])
cx, cy = rect[0]
box = cv.boxPoints(rect)
box = np.int0(box)
cv.drawContours(img_masked, [box], 0, (0, 0, 255), 2)
#cv.circle(img_masked, (np.int32(cx), np.int32(cy)), 2, (255, 0, 0), 2, 8, 0)
box_gray_color=[]
for by in range(box[2][1],box[0][1]+1):
for bx in range(box[1][0],box[3][0]+1):
#print(bx,by)
#cv.circle(img_masked,(bx, by), 1, (255, 0, 0), 2, 8, 0)
box_gray_color.append(gray[bx,by])
plt.hist(box_gray_color)
plt.hist(box_gray_color,bins=50)
plt.title(str(counter_number))
plt.show()
dist=cv.pointPolygonTest(cnts[i],(50,50),True)
"""
try:
cX = int(M["m10"] / M["m00"])
cY = int(M["m01"] / M["m00"])
cv.circle(img_masked, (cX, cY), 3, (255, 255, 255), -1)
cv.putText(img_masked, str(counter_number), (cX - 20, cY - 20),
cv.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 2)
area_of_cells_nucleus.append(cnt_area(cnts[i]))
location_cells_center[counter_number] = [cX, cY]
except:
pass
if counter_number == 1:
x1 = cX
y1 = cY
if counter_number == 2:
x2 = cX
y2 = cY
if counter_number == 1:
x_sample = cX
y_sample = cY
#cv.drawContours(img_masked, [cnts[i]], -1, (255, 255, 255), -1)#mask contours
print("Whole pic average cell nucleus area percent: ",
Whole_pic_cell_area_ave_percent)
print("Whole pic average cell nucleus area percent_ave: ",
numpy.mean(Whole_pic_cell_area_ave_percent))
print("Whole pic average cell nucleus color deep percent: ",
Whole_pic_cell_color_ave)
print("Whole pic average cell nucleus color deep percent_ave: ",
numpy.mean(Whole_pic_cell_color_ave))
cv.imshow('single_cell', img)
#-----put Text-----
print("total cells number : ", counter_number)
#cv.line(img_masked, (x1,y1), (x2,y2), (0,0,255), 2)
list_of_two_points = pixel_between_two_points(x1, x2, y1, y2)
#-----output information on the line
height_of_two_points = []
height_of_two_points_B = []
height_of_two_points_G = []
height_of_two_points_R = []
for m in range(0, len(list_of_two_points)):
height = img[list_of_two_points[m][1], list_of_two_points[m][0]]
try:
height_B = img[list_of_two_points[m][1],
list_of_two_points[m][0]][0]
height_G = img[list_of_two_points[m][1],
list_of_two_points[m][0]][1]
height_R = img[list_of_two_points[m][1],
list_of_two_points[m][0]][2]
height_of_two_points_B.append(height_B)
height_of_two_points_G.append(height_G)
height_of_two_points_R.append(height_R)
except:
pass
#print(height)
height_of_two_points.append(height)
img_sample = img.copy()
cv.circle(img_sample, (x_sample, y_sample), 3, (0, 0, 255), -1)
font = cv.FONT_HERSHEY_SIMPLEX
cv.putText(img_sample, "Sample_Point", (x_sample - 20, y_sample - 20),
font, 0.7, (255, 255, 255), 2)
#cv.imshow("img_sample_location_RED_DOT", img_sample)
# save to local
f = open("G:\\2020summer\\Project\\Cell_classfication_1.0.0\\dict.txt",
'w')
f.write(str(location_cells_center))
f.close()
# < list save
file1 = open('area_of_nucleus.txt', 'w')
for fp in area_of_cells_nucleus:
file1.write(str(fp))
file1.write('\n')
file1.close()
# list save >
cv.imwrite("G:\\2020summer\\Project\\Cell_classfication_1.0.0\\temp.bmp",
img_masked)
cv.imwrite("G:\\2020summer\\Project\\Cell_classfication_1.0.0\\temp_1.bmp",
img_nucleus_white_img)
cv.imwrite(
"G:\\2020summer\\Project\\Chromophobe_dataset1\\figure3_right.jpg",
img_masked)
#================hist of cells area==================
#plt.hist(cell_area_hist_list)
#plt.show()
#=================================
#-----
#=================================UI/
cv.putText(img_masked, "Overview", (80, 40), cv.FONT_HERSHEY_SIMPLEX, 1,
(0, 0, 0), 2)
image_size_text = "Image size: [Width :" + str(
img_original.shape[0]) + "]" + "[Height :" + str(
img_original.shape[1]) + "]"
cv.putText(img_masked, image_size_text, (80, img.shape[0] - 400),
cv.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 0), 2)
cv.putText(img_masked, "Total cells number: " + str(counter_number),
(80, img.shape[0] - 350), cv.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 0),
2)
cv.putText(img_masked, "Close window to continue",
(80, img.shape[0] - 250), cv.FONT_HERSHEY_SIMPLEX, 0.8,
(0, 0, 0), 1)
#=================================/UI
cv.imshow('img_copy', img_masked)
print("============Step 1 End============")
cv.waitKey()
return counter_number
def is_in_contour(cont, point):
x, y = point
return cv2.pointPolygonTest(cont, (x, y), False) > 0
for idx, (_curr, _next) in enumerate(zip(contours[::], contours[1::])):
# https://www.quora.com/How-do-I-iterate-through-a-list-in-python-while-comparing-the-values-at-adjacent-indices/answer/Jignasha-Patel-14
if article_complete:
article_mask = np.ones(
image.shape, dtype="uint8"
) * 255 # blank layer image for antother separate article
[cx, cy, cw, ch] = cv2.boundingRect(_curr)
[nx, ny, nw, nh] = cv2.boundingRect(_next)
if (ny - cy) > (nh + ch) * 2: # next is greater than current...
print('Big Gap! {}'.format(idx))
# loop through contents and insert any valid ones in this gap
for idxx in range(len(content_contours)):
[x, y, w, h] = cv2.boundingRect(content_contours[idxx])
# search_area_rect = cv2.rectangle(contents_mask,(cx,cy),(x+w,y+h),(0,0,255),thickness=3,shift=0)
dist = cv2.pointPolygonTest(content_contours[idxx], (x, y), False)
# https://stackoverflow.com/a/50670359/754432
if cy < y and cx - 10 < x and x < (
cx + w): # less than because it appears above
# check but not greater than the next title!!
if y > ny: # or next is another column
break
# cv2.drawContours(contents_mask, [c], -1, 0, -1)
# cv2.rectangle(contents_mask, (x,y), (x+w,y+h), (0, 0, 255), 3)
contents = contents_mask[y:y + h, x:x + w]
article_mask[
y:y + h, x:x +
w] = contents # copied title contour onto the blank image
image[y:y + h, x:x +
w] = 255 # nullified the title contour on original image
# cv2.putText(contents_mask, "#{},x{},y{}".format(idxx, x, y), cv2.boundingRect(contours[idxx])[:2], cv2.FONT_HERSHEY_PLAIN, 2.0, [255, 153, 255], 2) # [B, G, R]
def applyMask(
image_path,
mask_path,
save_path,
segmented_save_path,
mat_save,
threshold,
git_repo_base,
bregma_list,
atlas_to_brain_align,
model,
dlc_pts,
atlas_pts,
olfactory_check,
use_unet,
use_dlc,
plot_landmarks,
align_once,
atlas_label_list,
region_labels=True,
original_label=False,
):
"""
Use mask output from model to segment brain image into brain regions, and save various outputs.
:param image_path: path to folder where brain images are saved
:param mask_path: path to folder where masks are saved
:param save_path: path to overall folder for saving all images
:param segmented_save_path: path to overall folder for saving segmented/labelled brain images
:param mat_save: choose whether or not to output brain regions to .mat files
:param threshold: set threshold for segmentation of foregrounds
:param git_repo_base: The path to the base git repository containing necessary resources for MesoNet (reference
atlases, DeepLabCut config files, etc.)
:param bregma_list: The list of bregma locations (or landmarks closest to bregma).
:param region_labels: Choose whether or not to attempt to label each region with its name from the Allen Institute
Mouse Brain Atlas.
:param use_unet: Choose whether or not to define the borders of the cortex using a U-net model.
:param atlas_to_brain_align: If True, registers the atlas to each brain image. If False, registers each brain image
to the atlas.
:param model: The name of the U-net model (for passthrough to mask_functions.py)
:param dlc_pts: The landmarks for brain-atlas registration as determined by the DeepLabCut model.
:param atlas_pts: The landmarks for brain-atlas registration from the original brain atlas.
:param olfactory_check: If True, draws olfactory bulb contours on the brain image.
:param plot_landmarks: If True, plots DeepLabCut landmarks (large circles) and original alignment landmarks (small
circles) on final brain image.
:param atlas_label_list: A list of aligned atlases in which each brain region is filled with a unique numeric label.
This allows for consistent identification of brain regions across images. If original_label is True, this is an
empty list.
:param align_once: If True, carries out all alignments based on the alignment of the first atlas and brain. This can
save time if you have many frames of the same brain with a fixed camera position.
:param region_labels: choose whether to assign a name to each region based on an existing brain atlas (not currently
implemented).
:param original_label: If True, uses a brain region labelling approach that attempts to automatically sort brain
regions in a consistent order (left to right by hemisphere, then top to bottom for vertically aligned regions). This
approach may be more flexible if you're using a custom brain atlas (i.e. not one in which each region is filled with
a unique number).
"""
tif_list = glob.glob(os.path.join(image_path, "*tif"))
if atlas_to_brain_align:
if use_dlc and align_once:
image_name_arr = glob.glob(
os.path.join(mask_path, "*_brain_warp.png"))
else:
image_name_arr = glob.glob(os.path.join(image_path, "*.png"))
image_name_arr.sort(key=natural_sort_key)
if tif_list:
tif_stack = imageio.mimread(os.path.join(image_path, tif_list[0]))
image_name_arr = tif_stack
else:
# FOR ALIGNING BRAIN TO ATLAS
image_name_arr = glob.glob(os.path.join(mask_path, "*_brain_warp.png"))
image_name_arr.sort(key=natural_sort_key)
region_bgr_lower = (220, 220, 220)
region_bgr_upper = (255, 255, 255)
base_c_max = []
count = 0
# Find the contours of an existing set of brain regions (to be used to identify each new brain region by shape)
mat_files = glob.glob(
os.path.join(git_repo_base, "atlases/mat_contour_base/*.mat"))
mat_files.sort(key=natural_sort_key)
# adapt. from https://stackoverflow.com/questions/3016283/create-a-color-generator-from-given-colormap-in-matplotlib
cm = pylab.get_cmap("viridis")
colors = [cm(1.0 * i / NUM_COLORS)[0:3] for i in range(NUM_COLORS)]
colors = [
tuple(color_idx * 255 for color_idx in color_t) for color_t in colors
]
for file in mat_files:
mat = scipy.io.loadmat(
os.path.join(git_repo_base, "atlases/mat_contour_base/", file))
mat = mat["vect"]
ret, thresh = cv2.threshold(mat, 5, 255, cv2.THRESH_BINARY)
base_c = cv2.findContours(thresh, cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
base_c = imutils.grab_contours(base_c)
base_c_max.append(max(base_c, key=cv2.contourArea))
if not atlas_to_brain_align and use_unet:
# FOR ALIGNING ATLAS TO BRAIN
num_images = len(glob.glob(os.path.join(mask_path, "*_brain_warp*")))
output = os.path.join(mask_path, "..")
from mesonet.predict_regions import predictRegion
mask_generate = True
tif_list = glob.glob(os.path.join(image_path, "*tif"))
if tif_list:
input_path = image_path
else:
input_path = mask_path
predictRegion(
input_path,
num_images,
model,
output,
mat_save,
threshold,
mask_generate,
git_repo_base,
atlas_to_brain_align,
dlc_pts,
atlas_pts,
olfactory_check,
use_unet,
plot_landmarks,
align_once,
atlas_label_list,
region_labels,
original_label,
)
for i, item in enumerate(image_name_arr):
label_num = 0
if not atlas_to_brain_align:
atlas_path = os.path.join(mask_path, "{}_atlas.png".format(str(i)))
mask_input_path = os.path.join(mask_path, "{}.png".format(i))
mask_warped_path = os.path.join(
mask_path, "{}_mask_warped.png".format(str(i)))
atlas_to_mask(atlas_path, mask_input_path, mask_warped_path,
mask_path, i, use_unet, atlas_to_brain_align,
git_repo_base, olfactory_check, [])
new_data = []
if len(tif_list) != 0 and atlas_to_brain_align:
img = item
img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
else:
img = cv2.imread(item)
if atlas_to_brain_align:
img = cv2.resize(img, (512, 512))
if use_dlc:
bregma_x, bregma_y = bregma_list[i]
else:
bregma_x, bregma_y = [
round(img.shape[0] / 2),
round(img.shape[1] / 2)
]
original_label = True
mask = cv2.imread(os.path.join(mask_path, "{}.png".format(i)))
mask = cv2.resize(mask, (img.shape[0], img.shape[1]))
# Get the region of the mask that is white
mask_color = cv2.inRange(mask, region_bgr_lower, region_bgr_upper)
io.imsave(os.path.join(save_path, "{}_mask_binary.png".format(i)),
mask_color)
# Marker labelling
# noise removal
kernel = np.ones((3, 3), np.uint8) # 3, 3
mask_color = np.uint8(mask_color)
thresh_atlas, atlas_bw = cv2.threshold(mask_color, 128, 255, 0)
# if atlas_to_brain_align and use_dlc:
# atlas_bw = cv2.dilate(atlas_bw, kernel, iterations=1) # 1
# io.imsave(os.path.join(save_path, "{}_atlas_binary.png".format(i)), atlas_bw)
if not atlas_to_brain_align:
watershed_run_rule = i == 0
else:
if len(tif_list) == 0:
watershed_run_rule = True
else:
watershed_run_rule = i == 0
if align_once:
watershed_run_rule = i == 0
labels_from_region = []
if watershed_run_rule:
orig_list = []
orig_list_labels = []
orig_list_labels_left = []
orig_list_labels_right = []
# unique_regions = (np.unique(atlas_label)).tolist()
# unique_regions = [e for e in unique_regions if e.is_integer()]
unique_regions = [
-275,
-268,
-255,
-249,
-164,
-150,
-143,
-136,
-129,
-98,
-78,
-71,
-64,
-57,
-50,
-43,
-36,
-29,
-21,
-15,
0,
15,
21,
29,
36,
43,
50,
57,
64,
71,
78,
98,
129,
136,
143,
150,
164,
249,
255,
268,
275,
300,
400,
]
# atlas_label_df = pd.DataFrame(atlas_label)
# atlas_label_df.to_csv(os.path.join(save_path, "atlas_label.csv"))
cnts_orig = []
# Find contours in original aligned atlas
if atlas_to_brain_align and not original_label:
np.savetxt(
"atlas_label_list_{}.csv".format(i),
atlas_label_list[i],
delimiter=",",
)
for region_idx in unique_regions:
if region_idx in [300, 400]:
# workaround to address olfactory contours not being found
region = cv2.inRange(atlas_label_list[i],
region_idx - 5, region_idx + 5)
cnt_for_idx, hierarchy = cv2.findContours(
region.copy(), cv2.RETR_TREE,
cv2.CHAIN_APPROX_NONE)[-2:]
if len(cnt_for_idx) >= 1:
cnt_for_idx = cnt_for_idx[0]
else:
region = cv2.inRange(atlas_label_list[i], region_idx,
region_idx)
cnt_for_idx = cv2.findContours(region.copy(),
cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
cnt_for_idx = imutils.grab_contours(cnt_for_idx)
if len(cnt_for_idx) >= 1:
cnt_for_idx = max(cnt_for_idx, key=cv2.contourArea)
if len(cnt_for_idx) >= 1:
cnts_orig.append(cnt_for_idx)
labels_from_region.append(region_idx)
else:
cnts_orig = cv2.findContours(atlas_bw.copy(),
cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
cnts_orig = imutils.grab_contours(cnts_orig)
if not use_dlc:
# cnts_orig = cv2.findContours(
# atlas_bw.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE
#)
cnts_orig, hierarchy = cv2.findContours(
atlas_bw.copy(), cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)[-2:]
# cnts_orig = imutils.grab_contours(cnts_orig)
labels_cnts = []
for (num_label, cnt_orig) in enumerate(cnts_orig):
labels_cnts.append(cnt_orig)
try:
cv2.drawContours(img, cnt_orig, -1, (255, 0, 0), 1)
# io.imsave(os.path.join(segmented_save_path, "check_contour.png"), img)
except:
print("Could not draw contour!")
# try:
if atlas_to_brain_align:
c_orig_as_list = cnt_orig.tolist()
c_orig_as_list = [[c_val[0] for c_val in c_orig_as_list]]
else:
c_orig_as_list = cnt_orig.tolist()
c_orig_as_list = [[c_val[0] for c_val in c_orig_as_list]]
orig_polylabel = polylabel(c_orig_as_list)
orig_x, orig_y = int(orig_polylabel[0]), int(orig_polylabel[1])
if not original_label and atlas_to_brain_align:
label_to_use = unique_regions.index(
labels_from_region[num_label])
(text_width,
text_height) = cv2.getTextSize(str(label_to_use),
cv2.FONT_HERSHEY_SIMPLEX,
0.4,
thickness=1)[0]
label_jitter = 0
label_color = (0, 0, 255)
cv2.rectangle(
img,
(orig_x + label_jitter, orig_y + label_jitter),
(
orig_x + label_jitter + text_width,
orig_y + label_jitter - text_height,
),
(255, 255, 255),
cv2.FILLED,
)
cv2.putText(
img,
str(label_to_use),
(int(orig_x + label_jitter),
int(orig_y + label_jitter)),
cv2.FONT_HERSHEY_SIMPLEX,
0.4,
label_color,
1,
)
label_num += 1
orig_list.append((orig_x, orig_y))
orig_list_labels.append((orig_x - bregma_x, orig_y - bregma_y,
orig_x, orig_y, num_label))
if (orig_x - bregma_x) < 0:
orig_list_labels_left.append((
orig_x - bregma_x,
orig_y - bregma_y,
orig_x,
orig_y,
num_label,
))
elif (orig_x - bregma_x) > 0:
orig_list_labels_right.append((
orig_x - bregma_x,
orig_y - bregma_y,
orig_x,
orig_y,
num_label,
))
orig_list.sort()
orig_list_labels_sorted_left = sorted(orig_list_labels_left,
key=lambda t: t[0],
reverse=True)
orig_list_labels_sorted_right = sorted(orig_list_labels_right,
key=lambda t: t[0])
flatten = lambda l: [obj for sublist in l for obj in sublist]
orig_list_labels_sorted = flatten(
[orig_list_labels_sorted_left, orig_list_labels_sorted_right])
vertical_check = np.asarray(
[val[0] for val in orig_list_labels_sorted])
for (orig_coord_val,
orig_coord) in enumerate(orig_list_labels_sorted):
vertical_close = np.where(
(abs(vertical_check - orig_coord[0]) <= 5))
vertical_close_slice = vertical_close[0]
vertical_matches = np.asarray(
orig_list_labels_sorted)[vertical_close_slice]
if len(vertical_close_slice) > 1:
vertical_match_sorted = sorted(vertical_matches,
key=lambda t: t[1])
orig_list_labels_sorted_np = np.asarray(
orig_list_labels_sorted)
orig_list_labels_sorted_np[
vertical_close_slice] = vertical_match_sorted
orig_list_labels_sorted = orig_list_labels_sorted_np.tolist(
)
img = np.uint8(img)
else:
for num_label, cnt_orig in enumerate(cnts_orig): # cnts_orig
try:
cv2.drawContours(img, cnt_orig, -1, (255, 0, 0), 1)
except:
print("Could not draw contour!")
if not atlas_to_brain_align and use_unet:
cortex_mask = cv2.imread(
os.path.join(mask_path, "{}_mask.png".format(i)))
cortex_mask = cv2.cvtColor(cortex_mask, cv2.COLOR_RGB2GRAY)
thresh, cortex_mask_thresh = cv2.threshold(cortex_mask, 128, 255,
0)
cortex_cnt = cv2.findContours(cortex_mask_thresh,
cv2.RETR_EXTERNAL,
cv2.CHAIN_APPROX_NONE)
cortex_cnt = imutils.grab_contours(cortex_cnt)
cv2.drawContours(img, cortex_cnt, -1, (0, 0, 255), 3)
labels_x = []
labels_y = []
areas = []
sorted_labels_arr = []
label_jitter = 0
mask = np.zeros(mask_color.shape, dtype="uint8")
cnts = cnts_orig
print("LEN CNTS: {}".format(len(cnts)))
print("LEN LABELS: {}".format(len(orig_list_labels_sorted)))
if original_label or not atlas_to_brain_align:
labels_from_region = [0] * len(orig_list_labels_sorted)
for (z, cnt), (coord_idx, coord), label_from_region in zip(
enumerate(cnts), enumerate(orig_list_labels_sorted),
labels_from_region):
if atlas_to_brain_align and not original_label:
coord_label_num = unique_regions.index(
labels_from_region[coord_idx])
else:
coord_label_num = coord_idx
# compute the center of the contour
if len(cnts) > 1:
z = 0
c_x, c_y = int(coord[2]), int(coord[3])
c = cnt
if not atlas_to_brain_align and use_unet:
cnt_loc_label = ("inside" if [1.0] in [
list(
set([
cv2.pointPolygonTest(
cortex_sub_cnt,
(
c_coord.tolist()[0][0],
c_coord.tolist()[0][1],
),
False,
) for c_coord in c
])) for cortex_sub_cnt in cortex_cnt
] else "outside")
else:
cnt_loc_label = ""
rel_x = c_x - bregma_x
rel_y = c_y - bregma_y
pt_inside_cnt = [
coord_check for coord_check in orig_list_labels_sorted
if cv2.pointPolygonTest(c, (int(coord_check[2]),
int(coord_check[3])), False) == 1
]
if original_label:
try:
pt_inside_cnt_idx = orig_list_labels_sorted.index(
pt_inside_cnt[0])
label_for_mat = pt_inside_cnt_idx
except:
label_for_mat = coord_label_num
print(
"WARNING: label was not found in region. Order of labels may be incorrect!"
)
else:
label_for_mat = coord_label_num
# if cnt_loc_label != '':
# coord_label_num = "{} {}".format(coord_label_num, cnt_loc_label)
# The centroid of the contour works as the contour centre in most cases. However, sometimes the
# centroid is outside of the contour. As such, using the average x and y positions of the contour edges
# that intersect with the centroid could be a safer option. We try to find this average position and if
# there are more than two intersecting edges or if the average position is over 200 px from the
# centroid, we fall back to using the centroid as our measure of the centre of the contour.
# for coord in c_for_centre:
# if coord[0][0] == c_x:
# edge_coords_y.append(coord[0].tolist())
# if coord[0][1] == c_y:
# edge_coords_x.append(coord[0].tolist())
# print("{}: edge coords x: {}, edge coords y: {}".format(label, edge_coords_x, edge_coords_y))
# adj_centre_x = int(np.mean([edge_coords_x[0][0], edge_coords_x[-1][0]]))
# adj_centre_y = int(np.mean([edge_coords_y[0][1], edge_coords_y[-1][1]]))
# adj_centre = [adj_centre_x, adj_centre_y]
# if abs(adj_centre_x - c_x) <= 100 and abs(adj_centre_x - c_y) <= 100:
# print("adjusted centre: {}, {}".format(adj_centre[0], adj_centre[1]))
# c_x, c_y = (adj_centre[0], adj_centre[1])
# edge_coords_x = []
# edge_coords_y = []
# compute center relative to bregma
# rel_x = contour centre x coordinate - bregma x coordinate
# rel_y = contour centre y coordinate - bregma y coordinate
# print("Contour {}: centre ({}, {}), bregma ({}, {})".format(label, rel_x, rel_y, bregma_x, bregma_y))
c_rel_centre = [rel_x, rel_y]
if not os.path.isdir(
os.path.join(segmented_save_path, "mat_contour_centre")):
os.mkdir(
os.path.join(segmented_save_path, "mat_contour_centre"))
# If .mat save checkbox checked in GUI, save contour paths and centre to .mat files for each contour
if mat_save:
mat_save = True
else:
mat_save = False
# Prepares lists of the contours identified in the brain image, in the order that they are found by
# OpenCV
# labels_arr.append(label)
sorted_labels_arr.append(coord_label_num)
labels_x.append(int(c_x))
labels_y.append(int(c_y))
areas.append(cv2.contourArea(c))
# The first contour just outlines the entire image (which does not provide a useful label or .mat
# contour) so we'll ignore it
# if coord_label_num != 0:
# Cross-references each contour with a set of contours from a base brain atlas that was manually
# labelled with brain regions (as defined in 'region_labels.csv' in the 'atlases' folder). If the
# area of the contour is within 5000 square px of the original region and the centre of the contour
# is at most 100 px away from the centre of the original contour, label the contour with its
# corresponding brain region. Until we figure out how to consistently and accurately label small
# brain regions, we only label brain regions with an area greater than 1000 square px.
shape_list = []
label_color = (0, 0, 255)
for n_bc, bc in enumerate(base_c_max):
shape_compare = cv2.matchShapes(c, bc, 1, 0.0)
shape_list.append(shape_compare)
# for (n_r, r), (n_bc, bc) in zip(enumerate(regions.itertuples()), enumerate(base_c_max)):
# min_bc = list(bc[0][0])
# min_c = list(c[0][0])
# max_bc = list(bc[0][-1])
# max_c = list(c[0][-1])
#
# # 0.3, 75
# if label_num == 0 and region_labels and \
# (min(shape_list) - 0.3 <= cv2.matchShapes(c, bc, 1, 0.0) <= min(shape_list) + 0.3) and \
# min_bc[0] - 75 <= min_c[0] <= min_bc[0] + 75 and \
# min_bc[1] - 75 <= min_c[1] <= min_bc[1] + 75 and \
# max_bc[0] - 75 <= max_c[0] <= max_bc[0] + 75 and \
# max_bc[1] - 75 <= max_c[1] <= max_bc[1] + 75:
# # print("Current contour top left corner: {},{}".format(min_c[0], min_c[1]))
# # print("Baseline contour top left corner: {},{}".format(min_bc[0], min_bc[1]))
# closest_label = r.name
# cv2.putText(img, "{} ({})".format(closest_label, r.Index),
# (int(c_x + label_jitter), int(c_y + label_jitter)),
# cv2.FONT_HERSHEY_SIMPLEX, 0.3, label_color, 1)
# label_num += 1
# if label_num == 0 and not region_labels:
if (not region_labels
and original_label) or (not region_labels
and not atlas_to_brain_align):
(text_width,
text_height) = cv2.getTextSize(str(coord_label_num),
cv2.FONT_HERSHEY_SIMPLEX,
0.4,
thickness=1)[0]
cv2.rectangle(
img,
(c_x + label_jitter, c_y + label_jitter),
(c_x + label_jitter + text_width,
c_y + label_jitter - text_height),
(255, 255, 255),
cv2.FILLED,
)
cv2.putText(
img,
str(coord_label_num),
(int(c_x + label_jitter), int(c_y + label_jitter)),
cv2.FONT_HERSHEY_SIMPLEX,
0.4,
label_color,
1,
)
label_num += 1
if mat_save:
# Create an empty array of the same size as the contour, with the centre of the contour
# marked as "255"
c_total = np.zeros_like(mask)
c_centre = np.zeros_like(mask)
# Follow the path of the contour, setting every pixel along the path to 255
# Fill in the contour area with 1s
cv2.fillPoly(c_total, pts=[c], color=(255, 255, 255))
# Set the contour's centroid to 255
if c_x < mask.shape[0] and c_y < mask.shape[0]:
c_centre[c_x, c_y] = 255
if not os.path.isdir(
os.path.join(segmented_save_path, "mat_contour")):
os.mkdir(os.path.join(segmented_save_path, "mat_contour"))
if not os.path.isdir(
os.path.join(segmented_save_path,
"mat_contour_centre")):
os.mkdir(
os.path.join(segmented_save_path,
"mat_contour_centre"))
sio.savemat(
os.path.join(
segmented_save_path,
"mat_contour/roi_{}_{}_{}_{}.mat".format(
cnt_loc_label, i, label_for_mat, z),
),
{
"roi_{}_{}_{}_{}".format(cnt_loc_label, i, label_for_mat, z):
c_total
},
appendmat=False,
)
sio.savemat(
os.path.join(
segmented_save_path,
"mat_contour_centre/roi_centre_{}_{}_{}_{}.mat".format(
cnt_loc_label, i, label_for_mat, z),
),
{
"roi_centre_{}_{}_{}_{}".format(
cnt_loc_label, i, label_for_mat, z):
c_centre
},
appendmat=False,
)
sio.savemat(
os.path.join(
segmented_save_path,
"mat_contour_centre/rel_roi_centre_{}_{}_{}_{}.mat".
format(cnt_loc_label, i, label_for_mat, z),
),
{
"rel_roi_centre_{}_{}_{}_{}".format(
cnt_loc_label, i, label_for_mat, z):
c_rel_centre
},
appendmat=False,
)
count += 1
if align_once:
idx_to_use = 0
else:
idx_to_use = i
if plot_landmarks:
for pt, color in zip(dlc_pts[idx_to_use], colors):
cv2.circle(img, (int(pt[0]), int(pt[1])), 10, color, -1)
for pt, color in zip(atlas_pts[idx_to_use], colors):
cv2.circle(img, (int(pt[0]), int(pt[1])), 5, color, -1)
io.imsave(
os.path.join(segmented_save_path,
"{}_mask_segmented.png".format(i)), img)
img_edited = Image.open(
os.path.join(save_path, "{}_mask_binary.png".format(i)))
# Generates a transparent version of the brain atlas.
img_rgba = img_edited.convert("RGBA")
data = img_rgba.getdata()
for pixel in data:
if pixel[0] == 255 and pixel[1] == 255 and pixel[2] == 255:
new_data.append((pixel[0], pixel[1], pixel[2], 0))
else:
new_data.append(pixel)
img_rgba.putdata(new_data)
img_rgba.save(
os.path.join(save_path, "{}_mask_transparent.png".format(i)))
img_transparent = cv2.imread(
os.path.join(save_path, "{}_mask_transparent.png".format(i)))
img_trans_for_mat = np.uint8(img_transparent)
if mat_save:
sio.savemat(
os.path.join(segmented_save_path,
"mat_contour/transparent_{}".format(i)),
{"transparent_{}".format(i): img_trans_for_mat},
)
masked_img = cv2.bitwise_and(img, img_transparent, mask=mask_color)
if plot_landmarks:
for pt, color in zip(dlc_pts[idx_to_use], colors):
cv2.circle(masked_img, (int(pt[0]), int(pt[1])), 10, color, -1)
for pt, color in zip(atlas_pts[idx_to_use], colors):
cv2.circle(masked_img, (int(pt[0]), int(pt[1])), 5, color, -1)
io.imsave(os.path.join(save_path, "{}_overlay.png".format(i)),
masked_img)
print("Mask {} saved!".format(i))
d = {
"sorted_label": sorted_labels_arr,
"x": labels_x,
"y": labels_y,
"area": areas,
}
df = pd.DataFrame(data=d)
if not os.path.isdir(os.path.join(segmented_save_path,
"region_labels")):
os.mkdir(os.path.join(segmented_save_path, "region_labels"))
df.to_csv(
os.path.join(segmented_save_path, "region_labels",
"{}_region_labels.csv".format(i)))
print("Analysis complete! Check the outputs in the folders of {}.".format(
save_path))
k.clear_session()
if dlc_pts:
os.chdir("../..")
else:
os.chdir(os.path.join(save_path, '..'))
def isInContourV1(cont, pt, patch_size=None):
return 1 if cv2.pointPolygonTest(cont, pt, False) >= 0 else 0
def __init__(self, root, cropSize, trainFlag, imageRGBfilenames, rgbContours, rgbXYWHCR, depthImgFilename,
depthXYWHCR):
super(grantXdataset, self).__init__()
"""
Args:
root(string): directory with all the input images.
cropSize(integer): size of classification image (in meter)
trainFlag(string): 'train', 'vali' or 'test'
Outputs:
self.imageRGBfilenames: for future reading
self.rgbContours: The contours of all rgb pics
self.rgbXYWHCR: The originalX, originalY, pixel width, pixel height, cols and rows of rgb pics
self.trainShapes:
self.trainCategory:
self.testShapes:
self.testCategory:
self.valiShapes:
self.valiCategory:
"""
self.root = root
self.cropSize = cropSize
# read all rgb tif images and load the image's polygon for each image
self.imageRGBfilenames = imageRGBfilenames
self.rgbContours = rgbContours
self.rgbXYWHCR = rgbXYWHCR
self.depthImgFilename = depthImgFilename
self.depthXYWHCR = depthXYWHCR
self.trainFlag = trainFlag
# transform function
# self.toTensor = transforms.ToTensor()
# self.normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
# std=[0.229, 0.224, 0.225])
# augmentation
self.color_jitter = transforms.ColorJitter(0.1, 0.1, 0.1, 0.1)
# read all shapes and divide them
self.categories = {'Water': 0, 'Bog': 1, 'Channel_Fen': 2,
'Forest_Dense': 3, 'Forest_Sparse': 4, 'Wetland': 5,
'Dense_Forest': 3, 'Sparse_Forest': 4}
# training flag
if trainFlag == 'train':
self.augmentationFlag = True
else:
self.augmentationFlag = False
# sf = shapefile.Reader(join(root, "Class_Samples/Class_Samples.shp"))
sf = shapefile.Reader(join(self.root, "Class_Samples/{}Shapes.shp".format(trainFlag)))
shapes = []
shapeCategories = []
for rec in sf.shapeRecords():
pts = []
for point in rec.shape.points:
pts.append(point)
shapes.append(pts)
shapeCategories.append(self.categories[rec.record[-1]])
# training shape balance
if 'train' in trainFlag:
trainingShapeEachCategory = [[], [], [], [], [], []]
categoryState = np.zeros(6, dtype=np.int32)
for shape, category in zip(shapes, shapeCategories):
trainingShapeEachCategory[category].append(shape)
categoryState[category] += 1
upbound = np.max(categoryState)
for cntCategory, categoryNum in enumerate(categoryState):
for q in range(upbound - categoryNum):
trainingShapeEachCategory[cntCategory].append(
shapeAug(random.choice(trainingShapeEachCategory[cntCategory])))
shapes = []
shapeCategories = []
for cntCategory, shapeVec in enumerate(trainingShapeEachCategory):
for shape in shapeVec:
shapes.append(shape)
shapeCategories.append(cntCategory)
# writeShape(self.shapes, self.ShapeCategories, 'fortesting')
# generate RGB shapes for pics
self.originShapes = shapes
self.shapes = []
if cropSize is None:
pass
else:
for i, shape in enumerate(self.originShapes):
resizeRate = cropSize / 30
centerPt = (np.array(shape[0]) + np.array(shape[1]) + np.array(shape[2]) + np.array(shape[3])) / 4
newShape = []
for pt in shape:
newShape.append(tuple(resizeRate * (pt - centerPt) + centerPt))
self.shapes.append(newShape)
# generate which shape is in which image
shapeInImg = []
for i, shape in enumerate(self.shapes):
shapeInImg.append([])
for j, rgbCnt in enumerate(self.rgbContours):
for cnt, pt in enumerate(shape):
if cv2.pointPolygonTest(rgbCnt, pt, False) < 0:
break
if cnt == 3:
shapeInImg[i].append(j)
# if shape is not in any image
newOriginShape, newShape, newCategory, newShapeInImg = [], [], [], []
for i, shapeImgInfo in enumerate(shapeInImg):
if not shapeImgInfo:
continue
newOriginShape.append(self.originShapes[i])
newShape.append(self.shapes[i])
newCategory.append(shapeCategories[i])
newShapeInImg.append(shapeInImg[i])
self.originShapes = newOriginShape
self.shapes = newShape
self.ShapeCategories = newCategory
self.shapeInImg = newShapeInImg