基于多特征联合稀疏表示的SAR舰船目标识别方法

SAR Ship Target Recognition Method Using Joint Sparse Representation of Multiple Features

针对合成孔径雷达(Synthetic aperture radar,SAR)图像舰船目标识别问题,提出一种基于多特征联合稀疏表示的方法。分别采用几何尺寸特征矢量、主成分分析(Principal component analysis,PCA)和核主成分分析(Kernel PCA.KPCA)特征矢量描述SAR舰船目标的特性,基于以上特征,采用联合稀疏表示进行分类。根据散射特征的重构误差之和对测试样本的目标类别进行决策。实验中,采用RADARSAT-24类典型舰船目标的SAR图像进行性能测试。结果表明,此方法在标准操作条件下可以取得92.5%的平均识别率,高于现有的几类对比方法。另外,此方法在噪声干扰以及少量训练样本的条件下也能保持更强的稳健性,获得优于现有方法的性能。

To handle syntheti c aperture radar(SAR)ship target recognition problem,a method based on joint sparse representation of multiple features is proposed.The geometrical feature vector,principal component analysis(PCA)and kernel PCA(KPCA)feature vectors are utilized to describe the target characteristics of SAR ship targets respectively.Based on these above features,the joint sparse representation is used for classification.According to the sum of reconstruction errors of the scattering features,the target type of the tested sample is decided.In the experiments,SAR images of four typical ship targets measured by RADARSAT-2 are used for performance test,highter than the present several types of comparison method.According to the results,the proposed method could achieve an average recognition rate of 92.5%under the standard operating condition.In addition,under conditions like noise disturbance and few training samples,the proposed method could maintain stronger robustness and outperforms the compared ones.