Radar target recognition using contourlet packet transform and neural network approach

Contourlet is a “true” two-dimensional transform that captures the intrinsic geometrical structure and have been shown to be successful for many tasks in image processing. In this paper, a wavelet-based contourlet packet (WBCP) transform is investigated and an adaptive contourlet packet (ACP) transform based on genetic algorithm (GA) is proposed to extract the features of radar targets in synthetic aperture radar (SAR) images recognition. The features of the sampled targets are subsequently used to train a radical basis function neural network (RBFNN) that is then able to quickly and reliably recognize the objects. In comparison with WBCP, our proposed ACP has relatively low computational complexity and high recognition rate. Finally, we show some numerical experiments demonstrating the potential of this method for target recognition in SAR image processing.