Shape of corners array returned by findChessboardCorners()

user_4561314 0
Viewed 165

I'm using the OpenCV function findChessboardCorners() successfully, but I'm confused by the shape of the corners return value.

Here is my code below. I already know that my chessboard image has 8 x 6 internal corners.

import cv2

nx = 8
ny = 6

img = cv2.imread('test_image2.png')
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

ret, corners = cv2.findChessboardCorners(img, (nx, ny), None)
print(corners.shape)

The last line prints: (48, 1, 2).

Where does the 1 come from? I know there are 8 * 6 = 48 corners, and each corner has 2 coordinates. But what about the 1?

The documentation for findChessboardCorners() doesn't say anything about the corners result:

corners – Output array of detected corners.
opencv python
1 Answers

It is an unwanted and unnecessary dimension, you can eliminate the dimension by using the squeeze function of numpy:

 corners = np.squeeze(corners)

In this documentation, they have also mentioned in comment line as "extraneous singleton dimension" :

corners = np.squeeze(corners) # Get rid of extraneous singleton dimension

Edit:

I simply switched the code to C++ side to dig into. Lastly, here is the code and console output of it:

std::ccorners<<corners.rows<<"  "<<corners.cols<<"  "<<corners.dims<<std::endl;

Output:

48 1  2

So the output you are getting gives respectively:

  • rows number -> corner numbers
  • column number -> 1 as default because there is only one data to give
  • matrix dimensionality -> which is 2, number of data pair in each rows( coordinates )
user_110488870
Related