.png){kind=icon}
.png){kind=icon}
Return
A landslide identification method based on integrated segmentation network and transfer learning
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2025.131242.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Acquiring information on landslide locations and affected areas is crucial for geological disaster monitoring. This paper proposes a novel landslide identification method based on integrated segmentation network and transfer learning. Specifically, an integrated network architecture for landslide area segmentation is constructed using Kolmogorov-Arnold Network and Transformer-based U-Net, and the fusion of features at different scales is improved to accurately extract and restore the detailed features of landslides. To overcome the difficulty of landslide identification with indistinct optical features, a multi-modal image fusion network integrating Digital Elevation Model is proposed. A branch for extracting elevation features is added and a topographic-guided loss function is constructed. Furthermore, transfer learning is applied to improve the accuracy of landslide identification for small sample sizes. Experimental results validate the effectiveness of the proposed method in both single-modal identification of new landslides and multi-modal identification of old landslides.
Keywords:
landslide identification
integrated segmentation network
transfer learning
multi-modal image fusion
feature extraction