两向二维NMSD及其在SAR图像识别中的应用  

Discriminant Analysis of Two- Directional Two-Dimensional Nonparametric Maximum Scatter Difference and Its Application to SAR Target Recognition

作　　者：刘振[1] 姜晖[1] 秦立龙[1] 

机构地区：[1]电子工程学院,合肥230037

出　　处：《电子信息对抗技术》2013年第5期48-53,共6页Electronic Information Warfare Technology

摘　　要：结合非参数特征分析和最大散度差鉴别分析的思想,提出了两向二维非参数最大散度差((2D)2NMSD)鉴别分析,并用于SAR图像目标识别。首先计算二维图像的非参数散布矩阵,然后使用最大散度差准则求取投影矩阵,最后同时对数据图像矩阵的行方向和列方向进行特征提取。基于美国运动和静止目标获取与识别(MSTAR)公共数据库提供的实测数据的实验结果表明:该方法所提取的特征用于识别,可大大降低特征维数、提高识别性能,识别率可达98%以上。With the idea of nonparametfic feature analysis and maximum scatter difference discriminant analysis, the feature extraction method based on two-directional two dimensional nonparametric maxi- mum scatter difference （（2D）2NMSD） discfiminant analysis are proposed and applied to SAR （syn- thetic aperture radar） image target recognition. Firstly, nonparametric scatter matrix of two dimen- sional image matrix is computed. Then, projection vectors are computed using two dimensional maxi- mum scatter difference discfiminant analysis criterion. Finally, the feature matrix is extracted by ana- lyzing the row and column of image data at the same time. Experiments based on MSTAR （Moving and Stationary Target Acquisition and Recognition） public database demonstrate that using the feature extracted by （2D）2NMSD for recognition can improve the recognition performance with less feature di- mensionality and the recognition rate could be higher than 98 %.

关 键 词：线性鉴别分析 最大散度差鉴别分析 非参数特征分析 合成孔径雷达 目标识别 

分 类 号：TN957.52[电子电信—信号与信息处理]