Large Scene SAR Image Sample Construction Method Based on VAE_GAN Generated Data and Pixel Replacement

Aiming at the problem of high cost and difficulty in acquiring large scene SAR target image data, and the problem that it is difficult for intelligent generation algorithms to generate complex large scene images, this paper uses VAE_GAN to expand a small number of small scene SAR target images, and then uses the method of histogram matching to match the luminance histograms of the parts that need to be replaced in the SAR target image and the SAR scene image, and constructs a Gaussian image to smooth the matching result, and then use the pixel replacement method to fuse the processed target image with the scene image, and use the generated small scene target image and the existing SAR large scene background image to construct a large scene target image dataset. For VAE_GAN the introduction of the variational self-encoder is to enhance the diversity of samples generated by the generator; and the self-attention mechanism module is introduced to capture the long-distance dependence between location information.