C-RISE: A Post-Hoc Interpretation Method of Black-Box Models for SAR ATR

The integration of deep learning methods, especially Convolutional Neural Networks (CNN), and Synthetic Aperture Radar Automatic Target Recognition (SAR ATR) has been widely deployed in the field radar signal processing. Nevertheless, these methods are frequently regarded as black-box models due to limited visual interpretation their internal feature representation parameter organization. In this paper, we propose an innovative approach named C-RISE, which builds upon RISE algorithm provide a post-hoc technique for used SAR Images Recognition. C-RISE generates saliency maps that effectively visualize significance each pixel. Our outperforms by clustering masks capture similar fusion features into distinct groups, enabling more appropriate weight distribution increased focus on target area. Furthermore, employ Gaussian blur process masked area, preserving original image structure with optimal consistency integrity. extensively evaluated through experiments, results demonstrate superior performance over other based perturbation when applied neural networks recognition. our is highly robust transferable compared interpretable algorithms, including white-box methods.