With the global informationization boom, information security has become a problem for all of us. In order to be able to effectively detect the physical health status of criminals in prison and ensure the data security of the process, an image encryption method is designed to effectively protect the monitoring information. The process is based on generative adversarial network with generator and discriminator for image generation and data discrimination processing respectively, and optimizes the feature transmission process of image with the help of residual network. The key is generated by chaotic sequence method during the image transmission process. The encrypted image is transmitted to the staff port and the destination image is obtained after the decryption process of data key. The results of the study indicated that the decryption accuracy of the GAN algorithm in the dataset test increases gradually with the iteration process. The accuracy of the image after the completion of the iteration reached 98.69%, indicating that the algorithm has a good restoration effect for recovering the image after transmission. The structural similarity of the data image after the GAN algorithm processing decryption can reach 0.988. The peak signal-to-noise ratio index of the image was 37.78dB, which indicates that the clarity of the image after encrypted transmission is high. The research method can provide an effective theoretical support for the encrypted transmission of video images.