A Method for Reconstruction and Dynamic Prediction of Marine Water Pollution Scenarios Based on Generative Adversarial Networks

With the aggravation of marine water pollution, it is very important for marine environmental protection to accurately reconstruct the pollution scene and make a dynamic prediction. The purpose of this article is to develop an efficient and accurate method to solve the problems related to prediction and reconstruction of marine water pollution. By introducing generative adversarial network (i.e., generated countermeasure network) technology, a model of marine water pollution scene reconstruction and dynamic prediction based on this network was constructed. In the scene reconstruction experiment, after several rounds of training, structural similarity index measure was stabilized to 0.86, peak signal-to-noise ratio stabilized to around 28.7dB, and the structural integrity and clarity of the reconstructed scene were excellent. In the dynamic prediction experiment, mean square error was kept at a low level of 0.06 in a short time step, and the prediction trend is basically consistent with the actual observation.