针对合成孔径雷达(SAR)目标识别问题, 提出结合多特征联合表征和自适应加权的方法。分别采用主成分分析(PCA)、单演信号以及Zernike矩特征描述原始SAR图像, 获得3个对应的特征矢量。基于联合稀疏表示模型对3类特征进行联合表征。针对不同特征条件下的重构误差矢量, 采用自适应加权算法进行融合处理, 即在线性融合的框架下自适应获得权值, 达到良好的决策融合效果。最终, 根据融合后的误差对目标类别进行判定。实验基于MSTAR数据集开展, 针对10类目标识别问题分别在标准操作条件、噪声干扰和部分遮挡条件下进行测试, 结果验证了方法的有效性。

Aiming at the problem of synthetic aperture radar target recognition,a method combining multi-feature joint representation with adaptive weighting is proposed.The Principal Component Analysis (PCA), monogenic signal,and Zernike moment features are used to describe SAR images,and three corresponding feature vectors are obtained.Based on the joint sparse representation model,three corresponding features are jointly represented.The reconstruction error vectors from different features are fused using adaptive weighting algorithm under the framework of linear fusion.The optimal weights are achieved so the fused results can be improved.Finally,decision is made based on the fused reconstruction errors.Experiments are conducted on the MSTAR dataset for the 10-class problem under the standard operating condition,the conditions of noise corruption and partial occlusion,and the results verify the effectiveness of the method.