Dent detection after hail

Question

The following picture is a photo of hail damage on a car, taken with fringe projection.

I want to detect the dents and classify them (big ones in yellow and small ones in red).

How can I do that? Which algorithm / method do you suggest?

Answer 1

Thinning isn't a bad idea, but actually you don't really need it if you can binarize: the blob outlines will give equally good curves.

Follow every curve in turn and estimate the curvature, for instance by determining the deviation from a line segment between two points fifty pixels apart.

The sections of the curve with a significant deviation will tell you where damage is found.