OpenCV findContours循环Python

Question

I am having trouble getting my code to find the contours of the images that I create after the first one is created. The goal of this part of the program is to first create an image such as the following:

And then break it into images such as the following using the color_separator function. What this does is separate each individual image by color so that from the above image, we would get:

Then, I try to find the contours by using basic findContours in OpenCV. The problem lies in the findContours part: it will find the first set of contours; but for the others, the contours are blank. The images load up correctly; everything works up until findContours for all images after the first. So as an example, the green picture would have it's contours found and not print the "No Contours" check statement; but the rest of the images will print the "No Contours" check statement. If anyone can help, I'd really appreciate it.

def getAttributesFromNetwork(self,Network,image):
    ShapeList = []
    AngleList = []
    FillList = []
    SizeList = []
    print Network.letter
    # Open the image path from the Problems (Image Data) folder
    image = Image.open(image)
    grayscale = image.convert("L")
    blackwhite = grayscale.point(self.filter,"1")

    image = blackwhite
    image = image.convert("RGB")
    width,height = image.size
    colorindex = 0
    # Translate the image into pictures with various colors
    while True:
        color = DISTINCT_COLORS[colorindex]
        colorindex += 1
        blackpixel = None
        for x,y,pixel in self.walk(image):
            if pixel == (0,0,0):
                blackpixel = (x,y)
                break
        if not blackpixel:
            break

        neighbors = [blackpixel]
        while len(neighbors) > 0:
            processing = list(neighbors)
            neighbors = []
            for x,y in processing:
                image.putpixel((x,y),color)
                new = [(x-1,y),(x+1,y),(x,y-1),(x,y+1)]
                for x,y, in new:
                    if (x,y) in neighbors:
                        continue
                    if x < 0 or x >= width:
                        continue
                    if y < 0 or y >= height:
                        continue
                    if image.getpixel((x,y)) != (0,0,0):
                        continue
                    neighbors.append((x,y))
    # We use the count to save each network as a different image
    self.count = str(self.count)
    # Save the network image
    image.save("colored"+self.count+".png")
    # Open the network image; here, we'll convert it to a bunch of different 
    # images; each with a different shape
    im = Image.open("colored"+self.count+".png")
    # Separate the images
    colors_dict = color_separator(im)
    #print colors_dict
    # show the images:
    imageCount = 0
    # Iterate through the color dictionary for all of the images
    for key,value in colors_dict.iteritems():
        if key == (255, 255, 255):
            imageCount += 1
            continue
        imageCount = str(imageCount)
        # grab the individual image,
        image = value
        # save it,
        image.save(Network.letter+"coloredSmall"+imageCount+".png")
        # then read it back with OpenCV for processing
        img = cv2.imread(Network.letter+"coloredSmall"+imageCount+".png")
        # Convert it to grayscale; it processes better this way
        imgray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
        ret,thresh = cv2.threshold(imgray,127,255,0)
        # find the contours in the image
        contours,hierarchy = cv2.findContours(thresh,1,2)
        #count = 0
        # iterate through the contours,
        if not contours:
            print "No Contours"
        for cnt in contours:
            print "Looking through contours"
            #if (count%2) == 1:
                #count = count + 1
                #print "Count2: ",count
                #continue
            # approximate how many sides it has
            approx = cv2.approxPolyDP(cnt,0.01*cv2.arcLength(cnt,True),True)
            print len(approx)
            if len(approx) == 5:
                print "Half-Arrow"
                ShapeList.append("Half-Arrow")
            if len(approx) == 7:
                print "Arrow"
                ShapeList.append("Arrow")
            elif len(approx) == 3:
                print "Triangle"
                ShapeList.append("Triangle")
            elif len(approx) == 4:
                print "Square"
                ShapeList.append("Square")
            elif len(approx) >= 13:
                print "Circle"
                ShapeList.append("Circle")
            elif len(approx) == 12:
                print "Cross"
                ShapeList.append("Cross")
            (x,y),(MA,ma),angle = cv2.fitEllipse(cnt)
            AngleList.append(angle)
            #count = count + 1
            #print "Count: ",count3
            print ShapeList
        cv2.imshow("img",img)
        cv2.waitKey(0)
        cv2.destroyAllWindows()
        imageCount = int(imageCount)
        imageCount += 1

Answer 1

The problem was that the colors on the images were too dark when converted to grayscale. The green one was fine; but the dark blue did not convert well to grayscale at all. Moral of the story: When converting to grayscale from RGB, make sure your colors are light enough to convert.