摘要
局部保持投影(LPP)是一种能描述数据实际分布的流形学习算法,可以有效地捕获数据的局部信息。针对高精度SAR变形目标识别问题,文中提出一种结合L1图模型和LPP的SAR变形目标识别算法。考虑到稀疏描述具有判别力且对噪声具有鲁棒性的优点,构建L1图模型捕获样本之间的局部结构。此外,还采用一种正则化方法有效地解决了LPP算法中存在的矩阵奇异性问题。实测的MSTAR
The locality preserving projection(LPP)is a manifold learning algorithm that can describe the actual distribution of the data,and can effectively capture the local information of the data. In allusion to the recognition problem of high-precision synthetic aperture radar(SAR)morph targets,an SAR morph target recognition algorithm based on the combination of the L1-graph model and LPP is proposed in this paper. The L1-graph model is established to capture the local structure between samples considering that the sparse description has the discriminating power and noise robustness. A regularization method is adopted to effectively solve the matrix singularity problem of the LPP algorithm. The effectiveness of the proposed algorithm was verified by using the actually-measured MSTAR data.L1-graph modelSAR imagemorph target recognitionLPPsparse descriptionregularization
引文
[1] CLEMENTE C,PALLOTTA L,GAGLIONE D,et al. Automat?ic target recognition of military vehicles with Krawtchouk mo?ments[J]. IEEE transactions on aerospace and electronic sys?tems,2017,53(1):493?500.
[2] DONG G,KUANG G,WANG N,et al. Classification via sparserepresentation of steerable wavelet frames on Grassmann mani?fold:application to target recognition in SAR image[J]. IEEEtransactions on image processing,2017,26(6):2892?2904.
[3] LIU M,WU Y,ZHAO W,et al. Dempster?Shafer fusion ofmultiple sparse representations and statistical property for SARtarget configuration recognition[J]. IEEE geoscience and re?mote sensing letters,2014,11(6):1106?1110.
[4] LIU M,CHEN S. Label?dependent sparse representation for syn?thetic aperture radar target configuration recognition[J]. Interna?tional journal of remote sensing,2017,38(17):4868?4887.
[5] DING J,CHEN B,LIU H,et al. Convolutional neural networkwith data augmentation for SAR target recognition[J]. IEEEgeoscience and remote sensing letters,2016,13(3):364?368.
[6] DING B,WEN G,ZHONG J,et al. A robust similarity mea?sure for attributed scattering center sets with application toSAR ATR[J]. Neurocomputing,2017,219:130?143.
[7] AKBAR M,ALI A,AMIRA A,et al. An empirical study forPCA?and LDA?based feature reduction for gas identification[J]. IEEE sensors journal,2016,16(14):5734?5746.
[8] LIU M,WU Y,ZHANG Q,et al. Synthetic aperture radar tar?get configuration recognition using locality?preserving propertyand the Gamma distribution[J]. IET radar,sonar&naviga?tion,2016,10(2):256?263.
[9] LIU M,WU Y,ZHANG P,et al. SAR target configuration rec?ognition using locality preserving property and Gaussian mix?ture distribution[J]. IEEE geoscience and remote sensing let?ters,2013,10(2):268?272.
[10] DONG G,WANG N,KUANG G. Sparse representation ofmonogenic signal:with application to target recognition inSAR images[J]. IEEE signal processing letters,2014,21(8):952?956.
[11] MEYER C D. Matrix analysis and applied linear algebra[M].Cambridge:Cambridge University Press,2000.
[12] DAI D,YUEN P. Regularized discriminant analysis and itsapplication to face recognition[J]. Pattern recognition,2003,36:845?847.