A simple rotation invariant shape signature

Comparing shapes is a problem of both theoretical and practical importance in computer vision. A simple and robust solution to this problem could offer lots of potential applications in the food service industry/bakery environment. This paper highlights a machine vision based solution for measuring shapes for quality control purposes. We restrict ourselves to the study of the shape of plane objects in order not to deal with perspective corrections and we present a novel, fast rotation invariant shape descriptor developed for a shape machine inspection of biscuits. The paper discusses the overall algorithmic strategy of the quality control and examines the properties of a "good" shape signature. Finally, it deals with the histogram signature approach, presents some remarks on the adopted distance function and describes the testing environment and the experimental results.