TheCakeLover's profile - activity

Dear all,

I am constantly running into problems using convexityDefects(). I have created an example image (see below), and OpenCV correctly finds all contours and convex hulls, but still, when running the line of code I commented out it throws the error

error: (-215:Assertion failed) hpoints > 0 in function 'cv::convexityDefects'

Where is the error? Is it related to the image depth?

import cv2
import imutils
import numpy as np

black = np.zeros((500,500), dtype=np.uint8)
cv2.rectangle(black, (60,60), (440, 440), (255,255,255), 4)
poly = np.array([[80,80], [110, 110], [80, 200], [300, 110]])
cv2.drawContours(black, [poly], 0, (255, 255, 255), 3)
cnts = cv2.findContours(black, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
cnts = imutils.grab_contours(cnts)



for c in cnts:
    toShow = cv2.cvtColor(black.copy(), cv2.COLOR_GRAY2BGR)
    cv2.drawContours(toShow, [c], 0, (0,255,0), 2)
    hull = cv2.convexHull(c)
    cv2.drawContours(toShow, [hull], 0, (0,0,255), 1)

  # print("conv defects", cv2.convexityDefects(c, hull))

    cv2.imshow("cont", toShow)
    cv2.waitKey()

test image

inner contour with correct convex hull

drawContours() minimum line thickness Dear all, is there a way to reduce the thickness of drawContours() further than 0

Oh god, you can't imagine how often I read the documentation...thank you very much!

convexityDefects errors Dear all, I am constantly running into problems using convexityDefects(). I have created an exa

hpoints > 0: Error using convexityDefects() Dear all, I am grabbing the hull of a contour by using the convexHull()

Thank you very much! May I follow up with a slightly unrelated question? As I have feared, my code has to be wrong at so

Thank you very much! May I follow up with a slightly unrelated question? As I have feared, my code has to be wrong at so

I'm determining corner points using an approximated version of contour c, then try to find the index of the found corner in the original contour using numpy.where(). But for some points no index can be found, meaning the point seems not to be in the original contour.

As I understand the underlying DP-algorithm, no new points are added, is that correct?

Cheers!

Is the output of approxPolyDP(c) a subset of the input contour c? I'm determining corner points using an approximated ve