Single Camera Calibration APP question

I am trying to test the camera calibration app with some images. One image failed point detection because the APP should detect the 9x6 points at the center area but obviously the points on the very left and on the bottom are also detected. I have attached the image. Checkboard square is 50mm x 50mm. Please help me understand why this image fails. Thanks in advance.

[Matt J]

Why do you consider this a failure? Why shouldn't the edge corners be detected as well? Is the reprojection accuracy resulting from the calibration poor?

[Frank Qian]

Thanks Matt. This result is different from other images, such as the one below.

[Matt J]

Does it matter that they are different? In what way do you rely on them being the same?

Answer 1

[Dima Lisin]

Unfortunately, sometimes the checkerboard detetction algorithm fails. In this particular case, you can avoid this problem by using a checker board with wide white margins, or by placing it on light background.

[Frank Qian]

Thanks Dima. I will try a wide margin checkboard

[Frank Qian]

New image with wide margin board. Same result.

On the very right of the image, the edge corners are actually not the center of the green circles.

[Dima Lisin]

Hi Frank,

Does this happen for all your calibration images, or just some of them? As I said before, the checkerboard detection algorithm does not work 100% of the time. It typically works well enough to detect enough boards to calibrate.

Also, These appear to be synthetic images. What happens when you use real images?