Optimal projection space method for SAR target recognition

Automatic target recognition technique, based on the projections in the samples range space, can be availably used in SAR target recognition. It requires that the relativity between the range spaces of different targets should be minimal. However, the relativity between the corresponding range spaces is very big if the different targets have similar appearance and structure, which will considerably reduce the recognition performance. So, it is essential to construct new projection spaces from the range spaces of similar targets, so as to enlarge projection difference. In this paper, an efficient method for constructing the optimal projection space is proposed. The proposed method can effectively enlarge the difference between similar targets by the projections in the optimal projection space. Experimental results show that the optimal projection space can greatly increase the separability between different similar targets, and is superior to other classical methods in both recognition performance and efficiency.