Target Recognition of 3-D Synthetic Aperture Radar Images Via Deep Belief Network

Target recognition of 3-D synthetic aperture radar (SAR) images can obtain the image samples conveniently and quickly. This paper takes advantage of the image transformation between 3 dimensions and 2 dimensions to save the costs of constructing sample library. In the recognition method, meanwhile, deep belief network model has advantages over traditional recognition method, but the setup of network parameters ordinarily relies on experience. On the basis of deep belief network, this paper uses the cross-validation method (CV) to realize automatic optimization, and accomplish the recognition of 3-D SAR images. The simulation result shows that this method is superior to traditional recognition methods in both recognition rate and recognition time.