针对实际军事情况下车辆目标为非合作目标,提出改进的主成分分析方法(IPCA)。它首先利用稀疏求解方法得到与测试样本最相关的部分训练样本以及它们对测试样本的表示系数。然后结合主成分分析(PCA)得到最优投影矩阵,使投影后不同测试样本能更好地利用训练样本信息进行分类。利用美国运动和静止目标获取与识别数据库中3类目标进行识别实验,结果表明基于改进的PCA方法比传统的PCA方法能够得到更高的识别率,并对稀疏方位角训练样本有更好的鲁棒性。
In this work, an improved principal component analysis method (IPCA) is proposed for SAR target recognition. Firstly, we use the sparse method to obtain the training samples which are most relevant to the test samples and their representation coefficients for the test samples. Then, using the principal component analysis (PCA) we obtain the optimal projection matrix so that different test samples after projection can be better classified by using the training sample information. The results of experiments, performed on SAR ground stationary targets based on the moving and stationary target acquisition and recognition (MSTAR) database, show that IPCA reaches higher recognition rate and better robustness to sparse aspect training samples of three true objects than PCA.
[1] Maitre H.合成孔径雷达图像处理[M]. 孙洪,译. 北京:电子工业出版社,2005.
[2] Novak L M, Owirka G J, Brower W S. Performance of 10-and 20-target MSE classifiers[J]. IEEE Transactions on Aerospace and Electronic Systems, 2000, 36(4):1279-1289.
[3] Mishra A K, Motaung T. Application of linearand nonlinear PCA to SAR ATR[C]//25th Int-ernational Conference Radioelektronika, RADI-OELEKTRONIKA 2015. Pardubice, Czech Republic:the IEEE Inc, 2015:349-354.
[4] Cao Z, Min R, Pi Y, et al. The feasibility analysis of applying NMF in SAR target recognition[C]//IEEE International Conference on Digital Signal Processing, DSP 2015.Singapore:the IEEE Inc, 2015:721-725.
[5] 向卫力, 李晓辉, 周勇胜, 等. 一种鲁棒的多尺度稀疏表示SAR目标识别方法[J]. 中国科学院大学学报, 2016, 34(1):99-105.
[6] 刘中杰, 庄丽葵, 曹云峰, 等. 基于主元分析和稀疏表示的SAR图像目标识别[J]. 系统工程与电子技术, 2013, 35(2):282-286.
[7] 程建, 黎兰, 王海旭. 稀疏表示框架下的SAR目标识别[J]. 电子科技大学学报, 2014, 43(4):524-529.
[8] Thiagarajan J J, Ramamurthy K N, Knee P, et al. Sparse representations for automatic target classification in SAR images[C]//4th International Symposium on Communications, Control, and Signal Processing, ISCCSP-2010. Limassol, Cyprus:IEEE Computer Society, 2010:1-4.
[9] Morgan D A E. Deep convolutional neural networks for ATR from SAR imagery[C]//Algorithms for Synthetic Aperture Radar Imagery XXⅡ. Baltimore, MD, United states:SPIE, 2015:94750F.
[10] Wang H, Chen S, Xu F, et al. Application of deep-learning algorithms to MSTAR data[C]//IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2015. Milan, Italy:the IEEE Inc, 2015:3743-3745.
[11] Huan R H, Zhang P, Pan J. SAR ta-rget recognition using PCA, ICA and Gabor wavelet decision fusion[J]. Journal of Remote Sensing, 2012,16(2):268-274.
[12] 丁军, 刘宏伟, 陈渤, 等. 姿态图像缺失情况下的SAR目标识别[J]. 西安电子科技大学学报, 2016, 43(4):5-9.
[13] 徐勇, 范自柱, 张大鹏. 基于稀疏算法的人脸识别[M]. 北京:国防工业出版社, 2014.
[14] Candès E J, Tao T. The power of convex relaxation:Nearoptimal matrix completion[J]. IEEE Transactions on Information Theory, 2010, 56(5):2053-2080.
[15] Defense Advanced Research Project Agency (DARPA), Air Force Research Laboratory(AFRL).Moving and stationary target acquisition and recognition(MSTAR) public release dataset[DB/OL].(2015-09-21)[2016-10-20]. https://www.sdms.afrl.af.mil/datasets/mstar/.