Recognition Method of Landslide Remote Sensing Image based on EfficientNet

Due to the frequent occurrence of landslide disasters and its own complexity, landslide identification and disaster prevention become more difficult. Therefore, in-depth study of landslide rapid and accurate identification method is of great significance. In order to realize fast and precise landslide recognition in landslide remote sensing image, a landslide recognition method based on EfficientNet network extraction of landslide depth features was proposed in this paper. This method expands the neural network from three dimensions of depth, width and resolution by finding a group of optimal compound coefficients, and then adaptively optimizes the network structure. It introduces Stochastic Gradient Descent with Momentum (SGDM) algorithm as the optimizer of network learning. Fully considering the influence of the historical gradient, the algorithm adjusts the current gradient value constantly in the process of parameter updating. Accordingly, the parameter updating range of the network is adjusted constantly, so as to improve the learning effect of the neural network and extract the deep characteristics of the landslide. The experimental results show that the average accuracy of EfficientNet network model on the test set reaches 92.780 %. It can be more efficient and accurate for landslide identification, which is of great significance to the rapid response after the disaster.