Geometric Hashing Classifier Based on Modified D-S Theory in SAR Target Recognition

Geometric hashing technology can effectively recognize targets with partially changed shape. But when the known targets in training data set don´t satisfy with the condition of 360 azimuths, the effect of recognition degrades. The reason is that the imperfect training set and partially changed shape make the uncertain information increase. Dempster-Shafer theory can deal with the uncertainty successfully. However, Dempster-Shafer theory does not model well evidences with a high degree of conflict. In order to overcome the aforementioned problems, in this paper, a modified D-S theory combined with geometric hashing is proposed and used in SAR images recognition. Experimental results with MSTAR dataset show that this fusion method is feasible, and it can correctly recognize the targets with partially changed shape