基于多特征核协同表示的SAR 目标识别

为利用合成孔径雷达(synthetic aperture radar，SAR)目标不同特征数据间的相关性与互补性，提出一种基于多特征的Tikhonov 正则化核函数协同表示(multi-feature kernel collaborative representation- based classification withtikhonov regularization，MFKCRT)算法。采用美国运动和静止目标获取与识别(moving and stationary target acquisitionand recognition，MSTAR)计划公开发布的SAR 图像数据库进行实验，实现核函数变换空间上的多特征融合协同表示识别。实验结果表明：该算法相较于基本的协同表示，具有更优的可靠性与鲁棒性。

SAR Target Recognition Based on Multi-feature Kernel Cooperative Representation

In order to exploit the correlation and complementarity between different feature data of synthetic apertureradar (SAR) target, a Tikhonov regularization kernel function collaborative representation algorithm based on multiplefeatures is proposed. The SAR image database released by the moving and stationary target acquisition and recognition(MSTAR) program of the United States is used in the experiment. The multi-feature fusion collaborative representation andrecognition on the kernel function transformation space is realized. Experimental results show that the proposed algorithmis more reliable and robust than the basic collaborative representation.