Landslide area identification based on Dlv3plus-unet neural network

In order to accurately identify landslide geological hazards, a new model based on the combination of Deeplabv3+ and U-Net is proposed in this study. Firstly, data preprocessing was carried out on the remote sensing images of landslides occurring in Jiuzhaigou captured by Sichuan Mapping and Geographic Information Centre using UAVs, and each tif file was processed with channel number conversion. Secondly, the encoder layer of Deeplabv3+ is integrated into the decoder layer of U-Net to construct a deep-learning landslide recognition semantic segmentation model based on Dlv3plus-unet. Finally, the new model is compared with the mainstream semantic segmentation models, such as Deeplabv3+, U-Net, and Res-U-Net.The results show that Dlv3plus-unet has a better landslide recognition effect than Deeplabv3+, U-Net, and Res-U-Net, and is better than the other models in terms of the Accuracy, loss, and Jaccard index.