Two-Dimensional Maximum Clustering-Based Scatter Difference Discriminant Analysis for Synthetic Aperture Radar Automatic Target Recognition

Abstract

In this paper, a novel image feature extraction technique, called two-dimensional maximum clustering-based scatter difference (2DMCSD) discriminant analysis, is proposed. This method combines the ideas of two-dimensional clustering-based discriminant analysis (2DCDA) and maximum scatter difference (MSD), which can directly extract the optimal projection vectors from 2D image matrices rather than 1D image vectors based on the cluster scatter difference criterion. 2DMCSD not only avoids the linearity and singularity problems frequently occurred in the classical Fisher linear discriminant analysis (FLDA) due to the high dimensionality and small sample size problems, but also saves much time for feature extraction. Extensive experiments conducted on the moving and stationary target acquisition and recognition (MSTAR) public database demonstrate that the proposed method is more effective than the existing subspace analysis methods, such as two-dimensional principal component analysis (2DPCA) and two-dimensional linear discriminant analysis (2DLDA).