Object recognition by invariants

Deals with the interpretation and classification of optical images of three-dimensional scenes. An experimental system is described in which certain features, called Zernike moment invariants, are extracted from binary images and then used for automatic classification. Up to 20 features are used to describe binary images of five object types. The authors managed to achieve complete separation of the classes using 20 features. The Zernike invariants approach to the binary image analysis has proved to be a successful tool in object recognition.