子空间法雷达目标一维像识别研究

Subspace Methods of Radar Target Recognition Using Range Profiles

基于高分辨一维像,研究特征子空间法和正则子空间法在雷达目标识别中的应用。针对一维像敏感于目标姿态变化的特点,提出一种子空间串识别法,将所有姿态范围划分为一定数量的模区,在每模区建立各类目标的子空间.对未知目标,所处模区由雷达测定后,其一维像映射到该模区各类目标的特征子空间进行识别分类一单模区搜索准则.模拟和实测数据实验表明所提出方法是有效的.

Eigen-subspace and canonical-subspace methods are studied and applied to feature-extraction for target recognition using range profiles of a High-range-Resolution-Radar (HRR) system. Based on this study, a subspace cluster method is proposed to tackle the problem of aspect-sensitivity of range profiles. In subspace cluster method, the aspect scope of a radar target is divided into a proper number of zones, and eigen-subspaces are established for each zone. After the zone number of an unknown target is determined by radar, the range profile of this target is mapped into eigen-subspaces of the corresponding zone, and the class whose subspace has the maximum mapping energy is judged as the right class to which the unknown target belongs. This method is named as single-mode classification rule in the subspace cluster method. Experimental results on simulated data and field data show the efficiency of the subspace methods and subspace cluster method in target recognition.