Radar target recognition based on multiple features fusion with Dempster-Shafer theory

A radar high resolution range profile (HRRP) contains rather detailed structural information of a target and provides us a more reliable tool for target recognition. Various feature extraction methods have been developed and applied to radar target recognition successfully. Usually, different features contain different discriminative information. So the recognition performance will be improved by utilizing different features extracted from HRRP together. In this paper, a radar target recognition scheme based on multiple features fusion is presented. Firstly, four different features are obtained by using PCA, DCT, LDA and SNPE respectively. Then a simple nearest neighbor classifier (NNC) or nearest neighbor line classifier (NNLC) is employed for classification with each individual feature. Finally, based on the Dempster-Shafer evidential theory, the recognition results with each individual feature are fused efficiently to get a better recognition performance. Experimental results on measured profiles from three aircrafts indicate the effectiveness of the presented method.