Two-stage infrared image turbulence suppression method and system based on deep learning

The invention provides a two-stage infrared image turbulence suppression method and system based on deep learning. The method comprises the following steps: acquiring a to-be-recovered infrared turbulence image; inputting the to-be-recovered infrared turbulence image into a pre-trained turbulence removal model, and predicting to obtain a turbulence removal image; the pre-trained turbulence removing model comprises a first-stage deblurring model and a second-stage distortion removing model, the first-stage deblurring model is a U-shaped neural network based on Transform, feature extraction is performed on the infrared turbulence image to be recovered through a multi-head attention mechanism and a feedforward neural network on multiple scales, and a distortion image to be identified is obtained through mapping; and the second-stage distortion removal model performs channel-disrupted shuffling processing on the distorted image to be recognized to obtain a shuffled image to be recognized, and performs pixel super