 The proposed method is a novel approach to SAR image registration which uses a multiclass classification network to identify matching points between two SAR images. The network is trained using key points extracted from both images and then uses these points to generate sub-images which are used as training and testing data. This process helps to reduce the diversity between the two images, making it easier to find matching points. Additionally, a precise matching module is used to eliminate any inconsistencies in the results. Experiments show that the proposed method outperforms existing methods in terms of accuracy and robustness. This article was authored by Xiao Zhengdeng, Sha Sha Mao, Jin Yuan Yang, and others.