I have images of rocks of varied shape and texture, obtained by laser scanning (there are 300 images obtained by rotation). This means that the rock is effectively scanned in 360 degrees with slight rotation recorded by each in each image. There are 6 marks on the rock, circles, crosses, T, X's, and their combinations (an example is in the picture below). I want to automatically detect these shapes given a set of images (on at least 1/300 of them).
I've tried template matching with OpenCV, but I haven't had success. Do you know any methods or libraries that could solve this problem? I'm thinking about training a neural network for this.Example
