Abstract:When artificial intelligence technology is used for target recognition in the military field, in order to solve the problem of insufficient training data caused by the limitations of infrared image collection, a method of infrared image generation based on a generative adversarial network is proposed to achieve the expansion of the dataset. The basic generative adversarial network is improved, and the input of the network is changed from random noise to real images, so that the style conversion from image to image is realized, i.e. the color image is converted into an infrared image. After the model construction and training of the network, the experimental results show that the method can effectively generate clear and high-quality infrared images, and solve the problem of insufficient network training caused by insufficient infrared data.