How can I find the 2d-orientation of new images based on (several) training images?

Viewed 21

I have an image of a dartboard (with the numbers 1-20). If I rotate this image let's say 80 degrees, I want my python code to be able to compare the second image to the first one, and know that the angle of the second image is 80 degrees, compared to the first image. How should I start with the development of such a piece of code? Thank you!

1 Answers

Scikit-image is an amazing package with very good documentation. I would start by looking at the RANSAC algorithm.

In the above example, they use corner detection using skimage.feature.corner_harris.

My idea: If the dartboard image doesn't have any distinct corners, you may try any edge operators. Then apply RANSAC, and find the angle between the old coordinates and the new coordinates using basic trigonometry.

Related