Target recognition in SAR images via graph wavelet transform and 2DPCA

In this paper, we propose a new classification method based on graph wavelet filter banks and 2DPCA for target recognition in synthetic aperture radar (SAR) image. Graph wavelet transformation can provide multi-scale analysis similar to traditional wavelet transform, and it effectively detects image edge information in irregular domain. Therefore, the radar image is transformed into a graph wavelet domain using the Meyer spectral kernel function. Due to the two-dimensional PCA (2DPCA), developed from PCA, is common in pattern recognition and can extract features from two-dimensional SAR image directly, we introduce it to get features. The vectors derived from 2DPCA at different scales are then applied into the metasample-based sparse representation classifier (MSRC). Experiments on the moving and stationary acquisition and recognition (MSTAR) dataset demonstrate that the proposed method leads to an improvement in the recognition rate.