8

I have the following sample image:

enter image description here

I want to fill these triangles in the corners with white color. How could I detect them using OpenCV? Of course, in this particular sample, I can just lean on the gradient or brightness. Nevertheless, in the future images won't be such perfectly shaped, so I am thinking of some shape detection.

I heard that shaped can usually be detected with, for example, Hough transform. But I do not know what I should start with.

Contour detection in OpenCV does not help since it finds too many candidates. I tried to use approxPolyDP with size = 3, but there was also no result (there were no such objects found).

These triangles will always be triangles, but they do not need to touch bars every time. They always will be at the edges of the image. They share approximately the same area among them.

I would like to be able to detect triangles and collect points corresponding to these triangles in some container.

Improve this question
2
  • 3
    The errors in your contour detection are twofold: first, OpenCV finds the contours of the white parts of the image, so you'd need to invert it. But another problem stopping you even if you did invert it is that at least the bottom two triangles are connected to the bars. You state that you'll have harder images than these---it would probably be better to post your harder examples first. E.g. will they always be right triangles? Will they always be at the edges? What properties do these triangles have? Commented Sep 19, 2017 at 12:33
  • @AlexanderReynolds Unfortunately, I do not have harder images yet. These triangles will always be triangles, but they do not need to touch bars everytime. They always will be at the edges of image. They share approximately the same area. Commented Sep 19, 2017 at 12:50

1 Answer 1

Reset to default
12

I am able to detect the triangle with below code. I am finding all the contours in the image and then using approxPolyDP, I am able to find the triangle.

import cv2
import numpy as np
image_obj = cv2.imread('image.jpg')

gray = cv2.cvtColor(image_obj, cv2.COLOR_BGR2GRAY)

kernel = np.ones((4, 4), np.uint8)
dilation = cv2.dilate(gray, kernel, iterations=1)

blur = cv2.GaussianBlur(dilation, (5, 5), 0)


thresh = cv2.adaptiveThreshold(blur, 255, 1, 1, 11, 2)

# Now finding Contours         ###################
_, contours, _ = cv2.findContours(
    thresh, cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE)
coordinates = []
for cnt in contours:
        # [point_x, point_y, width, height] = cv2.boundingRect(cnt)
    approx = cv2.approxPolyDP(
        cnt, 0.07 * cv2.arcLength(cnt, True), True)
    if len(approx) == 3:
        coordinates.append([cnt])
        cv2.drawContours(image_obj, [cnt], 0, (0, 0, 255), 3)

cv2.imwrite("result.png", image_obj)

Output image enter image description here

You can get contours in coordinates list.

Improve this answer
Sign up to request clarification or add additional context in comments.

5 Comments

Thank you. But could you detect points that correspond to triangles? I mean collect all these points into some container.
I have modified the answer to get all the coordinates list.
@newt verify if the answer give expected results
By the way, it seems that OpenCV has some problems with correct recognition in the case if triangles boundaries touch bars. Do you know, maybe, how to improve this algorithm to detect triangles correctly in these cases? I mean, as you can see on your picture, there are parts of triangles which are not detected.
One approach can be to use erode. This way you can remove connecting edges. See here

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.