Application of Convolution Neural Network Based on Transfer Learning in Sandstorm Prediction in Inner Mongolia

There are six deserts and sandy lands in the central and western part of Inner Mongolia, which is one of the main sources of sandstorms in China. In most areas, the surface is dry, the precipitation is low and the wind is strong in winter and spring. The analysis and study of sandstorms in this area is of great significance to the study and prediction of sandstorms in China. Based on the comprehensive analysis of the research status of sandstorms at home and abroad, the application of convolution neural network to classify satellite cloud images and establish prediction models of sandstorms are relatively few. The convolution neural network algorithm based on transfer learning is used to classify infrared satellite cloud images to establish Sand-dust Storm Prediction model, and the prediction accuracy of sand-dust storm prediction model established under different learning strategies is compared in the paper. The results show that the model training speed is fast and the accuracy and generalization ability of the model are improved by using the parameter migration initialization network in transfer learning. The Sand-dust Storm Prediction Model Based on the convolution neural network algorithm of transfer learning has a higher accuracy in predicting the occurrence of sand-dust storm and the change of learning strategies has an important influence on the prediction accuracy of sandstorm prediction model.