def read_video_as_list_by_path(path, fname):
print 'reading video for ', path, fname
#cap = cv2.VideoCapture(path + '\\' + fname)
cap = cv2.VideoCapture()
cap.open(path + '\\' + fname)
fps = cap.get(cv2.cv.CV_CAP_PROP_FPS)
fourcc = cap.get(cv2.cv.CV_CAP_PROP_FOURCC)
width = cap.get(cv2.cv.CV_CAP_PROP_FRAME_WIDTH)
height = cap.get(cv2.cv.CV_CAP_PROP_FRAME_HEIGHT)
if not cap.isOpened():
print 'tools, read video as list by path : file not exist'
return None
f = []
while 1:
ret, frame = cap.read()
if ret is False:
break
f.append(frame)
if len(f) is 0:
print 'tools, read video as list by path : no frames in video'
return None
cap.release()
size = int2round((width, height))
fps = int2round(fps)
fourcc = int2round(fourcc)
return f, (fourcc, fps, size)
def get_number_interation_ransac(n_model, p_inliers, precision):
"""
x = 1 - ( 1 - pow(a,m) ) ^ n
x = log(1-p) / log(1 - pow(a,m)
:param n_model:
:param rate_inliers:
:param precision:
:return:
"""
lower = math.log(1 - math.pow(p_inliers, n_model), 10)
if lower == 0:
print 'my_ransac, get_iteration : lower is 0'
return 0
else:
x = math.log(1-precision, 10) / lower
return int2round(x)
def get_list_draw_every_ellipse(src, ellipse, vvpoints):
show_list = []
vvpoint = (int(vvpoints[0]), int(vvpoints[1]))
for i in range(len(src)):
img = src[i].copy()
if len(ellipse[i]) != 0:
for j in range(len(ellipse[i])): # draw all contours and line corresponding to contour.
cv2.ellipse(img, ellipse[i][j], (0, 255, 0), 1)
# ellipse_center = (int(ellipse[i][j][0][0]), int(ellipse[i][j][0][1]))
ellipse_center = int2round(ellipse[i][j][0])
cv2.line(img, vvpoint, ellipse_center, (0, 255, 0), 1, 8) # draw orientation line.
# a, b = get_end_points(ellipse[i][j], width, height)
# cv2.line(img, int2round(a), int2round(b), (255, 0, 0), 1) # draw line of ellipse
show_list.append(img)
return show_list
def preprocess_training_data(img, gt):
col = img.shape[0]
x = gt[0]
y = gt[1]
w = gt[2]
h = gt[3]
o = gt[4]
center = (x + (w / 2), y + (h / 2))
distance = get_distance_two_points(center, (x, y))
distance = int2round(distance)
if center[1] - distance < 0:
return None
if center[1] + distance > col:
return None
if len(img) == 2:
roi = img[center[1] - distance: center[1] + distance, x: x + w]
else:
roi = img[center[1] - distance: center[1] + distance, x: x + w, :]
rotate_matrix = cv2.getRotationMatrix2D((w / 2, distance), o, 1) # TODO check angle system.
return cv2.warpAffine(roi, rotate_matrix, (w, 2 * distance)) # , flags=cv2.INTER_LINEAR)
def prepare_size_ransac(self, ellipse_list):
# ellipse_table = [[[[], []] for x in range(shape[1])] for x in range(shape[0])] # repeat [[row] for col]
size_dic = {}
for i in range(len(ellipse_list)): # assgning ellipse at correponding point
e = ellipse_list[i]
center = int2round(e[0])
size = e[1]
dist = self.get_distance_two_points(self.vvpoint, center)
if dist not in size_dic:
size_dic[dist] = [[], []]
size_dic[dist][0].append(size[0]) # shor axis = width
size_dic[dist][1].append(size[1]) # long axis = height
result = {}
for key in size_dic.keys():
height = np.array(size_dic[key][1])
width = np.array(size_dic[key][0])
result[key] = (width.mean(), height.mean())
# return height_list,width_list
return result
def test_size(self, ellipse_list):
base_list = []
height_list = []
width_list = []
size_dic = {}
given_vvpoint = (584, 144)
for i in range(len(ellipse_list)): # assgning ellipse at correponding point
e = ellipse_list[i]
center = int2round(e[0])
size = e[1]
# dist = self.get_distance_two_points(self.vvpoint, center)
dist = self.get_distance_two_points(given_vvpoint, center)
base_list.append(dist)
width_list.append(size[0])
height_list.append(size[1])
if dist not in size_dic:
size_dic[dist] = [[], []]
size_dic[dist][0].append(size[0]) # shor axis = width
size_dic[dist][1].append(size[1]) # long axis = height
plot2d([base_list, width_list, height_list], ["distance", "raw_width", "raw_height"], False, True)
print len(base_list), len(width_list), len(height_list)
height_mean_list = []
width_mean_list = []
for key in size_dic.keys():
height = np.array(size_dic[key][1])
width = np.array(size_dic[key][0])
height_mean_list.append((height.mean()))
width_mean_list.append(width.mean())
print len(size_dic.keys()), len(height_mean_list)
plot2d(
[size_dic.keys(), width_mean_list, height_mean_list], ["distance", "mean width", "mean height"], False, True
)