摘要
针对传统KPCA方法识别SAR图像时需首先对图像进行拉直处理问题进行研究,借助于图像欧式距离,提出了一种基于图像欧式距离的核函数构建方法,给出了3种基于图像欧式距离的核函数,并将其应用SAR图像识别中。以实测MSTAR数据为例,给出了识别结果和仿真分析,结果证明了该算法能够有效克服传统KPCA方法的局限性,是一种可行的方法。
Using KPCA for SAR image recognition,the image is firstly transformed into 1D vectors,which makes structure information of 2D image lost and the dimensions of 1D vectors very high.In this paper,an image Euclidean Distance KPCA method is presented,and three kernel functions based on image Euclidean Distance are also proposed.Examples of recognition MSTAR dataset are selected to demonstrate this new method.And experimental results show that the methos is effective and feasible.
引文
[1]张德祥,吴小培,高清维,等.基于平稳Contourlet变换的极化SAR图像融合[J].电子科技大学学报,2010,39(2):200-203.
[2]Hichem S.Kernel PCA for Similarity Invariant Shape Recognition[J].Neural Computing,2006(10):1-12.
[3]Nikolaos N,Adrian G B.Kernel-Based Classification using Quantum Mechanics[J].Pattern Recognitio,2007(40):875-889.
[4]张友静,黄浩.基于KPCA和SAM的城市植被遥感分类研究[J].地理与地理信息科学,2006,22(3),35-38.
[5]Zhao Q,Jose C.Support Machines for Automatic Electronic TargetRecognition[J].IEEE Vector Trans.on Aerospace SAR and Ystems,2001,37(2).643-654.
[6]韩萍,吴仁彪.基于KPCA准则的SAR目标特征提取与识别[J].电子与信息学报,2003,25(10):1297-1301.
[7]Wang L W,Zhang Y,Feng J F.On the Euclidean Distance of Images[J].IEEE Trans,Pattern Anal,Mach,Intell,2005,27(8):1334-1339.
[8]陈渤,刘宏伟,保铮,等.一种针对雷达高分辨距离像的融合核优化算法[J].电子学报,2006:34(6):1146-1151.