| 摘要: |
| 在分析传统Fisher线性鉴别分析局限性的基础上，由图像的行信息和列信息提出了两
种形式的二维非参数特征分析 (2DNFA)的特征提取方法，并应用于SAR图像目标的识别。直
接在SAR图像矩阵上使用非参数特征分析提取特征不仅能充分发挥非参数特征分析的性能而
且保留了图像矩阵的结构信息，大大降低了散度矩阵的维数，减小了运算量。使用美国MSTA
R计划录取的数据对算法进行了仿真验证，实验结果显示两种形式的二维非参数特征分析在
较低特征维数下的识别率均可以达到98%以上，表明所提方法的有效性和正确性。 |
| 关键词: 合成孔径雷达 目标识别 Fisher线性鉴别分析 非参数特征分析 |
| DOI： |
|
| 基金项目: |
|
| SAR image target recognition based on two dimensional nonparametric feature analysis |
| LIU Zhen,JIANG Hui,XU Hai-feng |
| () |
| Abstract: |
| Based on the analysis of traditional Fisher Linear Discriminant Analys
is（FLDA）, two forms of feature extraction methods of two dimensional Nonparame
tric Feature Analysis(NFA) based on the image′s row information and column information
are proposed and also applied to SAR(Synthetic Aperture Radar) image target recogniti
on. The method that extracts feature by using nonparametric feature analysis d
irectly on the SAR image matrix can not only give full play to the performance of
nonparametric feature analysis but also retain the structural information of th
e image matrix, and greatly reduce the dimension of the scatter matrix and the c
omputing complexity. Experiments on MSTAR (Moving and Stationary Target Acquisition
and Recognition) public database show that recognition rates of the new methods
can achieve more than 98% with less feature dimensionality, indicating the corre
ctness and effectiveness of the proposed methods. |
| Key words: SAR target recognition Fisher linear discriminant analysis nonparametric feature analysis |
《电讯技术》编辑部 