SAR Image Target Recognition Based on Two Dimensional Nonparametric Maximum Scatter Difference Discriminant Analysis

Hui JIANG,Zhen LIU,Peng WANG
DOI: https://doi.org/10.3969/j.issn.1002-0640.2014.05.025
2014-01-01
Abstract:To enhance the performance of linear discriminant analysis and avoid its“small sample problem”,an image feature extraction method based on two dimensional nonparametric maximum scatter difference(2DNMSD)discriminant analysis is proposed. In this method,the scatter matrices are constructed directly on the SAR image matrices by non-parametric feature analysis criterion. Then, projection vectors are computed using two dimensional maximum scatter difference discriminant analysis criterion. Experiments on MSTAR(Moving and Stationary Target Acquisition and Recognition) public database show that recognition rates of the new method can achieve more than 98 % with less feature dimensionality,which indicates the correctness and effectiveness of our method.
What problem does this paper attempt to address?