A SVR-based SAR target azimuth fusion estimation

In SAR ATR, target azimuth estimation is one of the important pre-processing steps. Accurate estimation will bring a higher recognition rate. This paper presents a novel target azimuth estimation algorithm based on SVR with feature fusion. First target imaging pose is estimated roughly by long-edge fitting (LF) and ridgelet transform (RT) respectively, then SVR estimators are trained by the above estimated results which are combined together as the sample features. With the SVR, SAR target azimuth can be estimated more accurately. Experimental results with Moving and Stationary Target Acquisition and Recognition (MSTAR) database show the proposed method is effective, and can give better estimation performance than that either by LF or RT.