Shape recognition using simple measures of projections

The development of an efficient global shape-recognition scheme is considered. An algorithm based on simple measures of projections has been developed to classify shapes. An object is characterized by its projections at various angles. Some statistical features are identified to represent the projections and are related to the moments of the object. Feature vectors are computed from these relations and used to identify a shape against a library of known shapes. The method is tested on two-dimensional shapes, and the results are presented. On the basis of these results, the merits of the method are discussed.<>