E-DeepLabV3+: A Landslide Detection Method for Remote Sensing Images

Abstract

The timely detection after landslides is significant for disaster mitigation and disaster reconstruction. In the application of landslide detection for remote sensing images, the semantic segmentation network DeepLabV3+ contains too many parameters, and the detection results are not ideal. This paper proposes a network model E-DeepLabV3+ for landslide detection in remote sensing images in response to the above problems. This model takes EfficientNet as the backbone network, the Matthews Correlation Coefficient as the loss function, and adjusts the learning rate through cosine annealing decay. The experimental results on the Bijie landslide dataset show that, compared with DeepLabV3+, E-DeepLabV3+ significantly reduces the number of parameters, which is more conducive to training and deployment. At the same time, it can provide better landslide detection results and effectively delineate the scope of landslides.