Abstract: The aim of this research is to enhance the extraction accuracy by improving the classification of micro-terrain landslide remote sensing information. The landslide information in local areas of Yichang was extracted by using the method of Convolutional Neural Networks (CNN) combined with Convolutional Block Attention Module (CBAM) based on the fusion of Unmanned Aerial Vehicle (UAV) hyperspectral image (HSI) and Light Detection and Ranging (LiDAR) data. Results demonstrated that landslide information can be extracted with more accuracy based on the advantages of hyperspectral and radar data.

Key words: hyperspectral image, LiDAR, data fusion, CBAM, landslide information extraction