Deep convolutional GAN-based optical image and SAR image registration method

The invention discloses a deep convolutional GAN-based optical image and SAR image registration method. The method comprises the following steps of: obtaining a training sample; constructing two generative adversarial networks; expanding training sample data; constructing a feature extraction and matching network; training the feature extraction and matching network by adoption of a cross iteration strategy; predicting a matching relationship; removing mis-matching points; calculating a geometric transformation matrix; and registering images. According to the method, the problems that neural network training samples are lack of data, samples are single and extracted image features lose space information in the prior art are overcome, the robustness of different-source image registration iseffectively improved, and higher-precision registration for SAR images and optical images is realized.