def parseImageWithMask(img, mask, show, name):
target = img.copy()
image_utils.applyMask(target, mask)
target = cv2.cvtColor(target, cv2.COLOR_BGR2GRAY)
colorThreshold = 40
notBlack = numpy.sum(target > 0)
countAbove = numpy.sum(target > colorThreshold)
countBelow = notBlack - countAbove
#for rows in target:
# for pixel in rows:
# if pixel != 0:
# if pixel > colorThreshold:
# countAbove += 1
# else:
# countBelow += 1
stau = countBelow > countAbove
info = {}
info["jam"] = stau
info["vehicles"] = 0
return (info, target)
def validateSingle(image, mask, info, name):
img = image.copy()
image_utils.applyMask(img, mask)
text = "Kein Stau"
color = (0, 255, 0)
if info["jam"]:
text = "Stau"
color = (0, 0, 255)
stau = True
cv2.putText(img, text, (10, 50), cv2.FONT_HERSHEY_SIMPLEX, 1, color, 1,
cv2.LINE_AA)
cv2.imshow(name, img)
key = cv2.waitKey(0) & 0xFF
cv2.destroyAllWindows()
if key == 13: # Enter - Stau
print("Stau")
return True
print("Kein Stau")
return False
def parseImageWithMask(img, mask, show, name):
target = img.copy()
med = cv2.GaussianBlur(target, (5, 5), 1.2)
image_utils.applyMask(med, mask)
med = cv2.cvtColor(med, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(med, 100, 200)
#cv2.imshow("hi1", edges)
#cv2.waitKey(0)
fgthres = cv2.threshold(edges.copy(), 200, 255, cv2.THRESH_BINARY)[1]
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (2, 2))
closing = cv2.morphologyEx(fgthres, cv2.MORPH_CLOSE, kernel)
opening = cv2.morphologyEx(closing, cv2.MORPH_OPEN, kernel)
dilation = cv2.dilate(opening, kernel, iterations=2)
img_dilation = dilation
vehicles = image_utils.findVehicles(img_dilation)
stau = len(vehicles) > 10
info = {}
info["jam"] = stau
info["vehicles"] = 0
return (info, target)
def parseImageWithMask(img, mask, show, name, classifier):
target = img.copy()
image_utils.applyMask(target, mask)
target = cv2.cvtColor(target, cv2.COLOR_BGR2GRAY)
cars = classifier.detectMultiScale(target, 1.1, 1)
stau = len(cars) > 10
info = {}
info["jam"] = stau
info["vehicles"] = 0
return (info, target)
def parseImageWithMask(img, bgs, mask, show, name):
target = img.copy()
med = cv2.GaussianBlur(target, (5, 5), 1.2)
image_utils.applyMask(med, mask)
image_utils.applyMask(target, mask)
image = bgs.apply(med)
fgthres = cv2.threshold(image.copy(), 200, 255, cv2.THRESH_BINARY)[1]
kernel = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (2, 2))
closing = cv2.morphologyEx(fgthres, cv2.MORPH_CLOSE, kernel)
opening = cv2.morphologyEx(closing, cv2.MORPH_OPEN, kernel)
dilation = cv2.dilate(opening, kernel, iterations=2)
img_dilation = dilation
vehicles = image_utils.findVehicles(img_dilation)
for vehicle in vehicles:
cv2.rectangle(target, (vehicle[0], vehicle[1]), (vehicle[0] +
vehicle[2], vehicle[1] + vehicle[3]), (0, 255, 0))
text = "Kein Stau"
color = (0, 255, 0)
stau = False
if len(vehicles) >= 10:
text = "Stau"
color = (0,0,255)
stau = True
cv2.putText(target,text,(10,100), cv2.FONT_HERSHEY_SIMPLEX, 2,color,2,cv2.LINE_AA)
if show:
cv2.imshow(name, target)
info = {}
info["jam"] = stau
info["vehicles"] = len(vehicles)
return (info, target)
def train_image(pic1, pic2, alpha):
pf1 = PointFinder(pic1)
pf2 = PointFinder(pic2)
tri1, landmarks, image, convex = getTriangulation(pf1)
tri2, landmarks2, image1, convex = getTriangulation(pf2)
for point in convex:
print(point)
b, g, r = cv2.split(image1)
outputImage = np.zeros((image.shape[0], image.shape[1], 4), np.uint8)
outputImage[:, :, 0] = b
outputImage[:, :, 1] = g
outputImage[:, :, 2] = r
outputImage[:, :, 3] = 255
count = 0
landmarks3 = []
for i in range(0, len(landmarks)):
x = (1 - alpha) * landmarks2[i][0] + alpha * landmarks2[i][0]
y = (1 - alpha) * landmarks2[i][1] + alpha * landmarks2[i][1]
landmarks3.append((x, y))
for tri in tri1:
first = np.float32(
[landmarks[tri[0]], landmarks[tri[1]], landmarks[tri[2]]])
second = np.float32(
[landmarks3[tri[0]], landmarks3[tri[1]], landmarks3[tri[2]]])
r1 = cv2.boundingRect(first)
r2 = cv2.boundingRect(second)
(x, y, w, h) = r1
(x1, y1, w1, h1) = r2
firstCropped = np.array([[first[0][0] - x, first[0][1] - y],
[first[1][0] - x, first[1][1] - y],
[first[2][0] - x, first[2][1] - y]], np.int32)
secondCropped = np.array([[second[0][0] - x1, second[0][1] - y1],
[second[1][0] - x1, second[1][1] - y1],
[second[2][0] - x1, second[2][1] - y1]],
np.int32)
try:
affine = cv2.getAffineTransform(np.float32(firstCropped),
np.float32(secondCropped))
except TypeError as t:
print("errored on")
print(type(first))
print(type(second))
# warp triangle from first image to shape of second
cropped = image[y:y + h, x:x + w]
mask = np.zeros_like(cropped)
cv2.fillConvexPoly(mask, firstCropped, 255)
masked = applyMask(cropped, mask)
warped = cv2.warpAffine(masked, affine, (w1, h1))
#add triangle to full image
area = outputImage[y1:y1 + h1, x1:x1 + w1]
overlay(area, warped)
outputImage[y1:y1 + h1, x1:x1 + w1] = area
count += 1
return outputImage