An effective regional descriptor and its application to target recognition

Abstract

A novel effective regional descriptor based on the normalized rapid descriptor (NRD) and regional projections is presented. The NRD is significantly superior to the normalized Fourier descriptor (NFD) in computational speed and storage requirement. The proposed descriptor was used for shape analytical experiments for several types of synthesized images. The descriptor was applied to the target recognition of four sequential images varying in size and orientation relative to the sensor. The results show that the descriptor is quite effective.<>