Research on Prediction Algorithm of Satellite Nephogram Based on GAN

The traditional work in the field of nephogram prediction is largely based on the premise that the cloud system remains stable. These methods have poor performance in predicting cloud trends, and cannot meet the effectiveness of prediction, which greatly limits the availability of cloud prediction. There are two main challenges in nephogram prediction. Firstly, most of the observed cloud changes are non-stationary and non-linear irregular changes. Secondly, different cloud systems change in the same region are also different, so it is difficult to establish an efficient model for description. In view of the above problems, this paper uses the improved generative adversarial network to predict the nephogram, and predicts the future frames of the nephogram according to the input frame sequence. This method abandons the complex mathematical operation and the premise of stability and linearity. Using enough data and fewer constraints to the model. The experimental results on FY-4 nephogram data show that the experimental results of using this method to predict satellite nephogram are better than those predicted by traditional methods and general GAN model.