An object-based spectral and elevation feature fusion framework for landslide mapping using time-series Landsat-8 imagery

This study presents an object-based spectral and elevation feature fusion framework for landslide mapping using time-series Landsat-8 imagery. The proposed approach integrates time-series multispectral image data with digital elevation models to enhance the robustness and accuracy of landslide detec...

Full description

Saved in:
Bibliographic Details
Main Authors: Tsung-Han Wen, Tee-Ann Teo
Format: Article
Language:English
Published: Taylor & Francis Group 2025-12-01
Series:Geomatics, Natural Hazards & Risk
Subjects:
Online Access:https://www.tandfonline.com/doi/10.1080/19475705.2025.2471014
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This study presents an object-based spectral and elevation feature fusion framework for landslide mapping using time-series Landsat-8 imagery. The proposed approach integrates time-series multispectral image data with digital elevation models to enhance the robustness and accuracy of landslide detection. The methodology employs multiresolution segmentation with a fusion of spectral and topographic features, enabling the model to capture complex patterns and improve segmentation quality. The study used the multivariate long short-term memory fully convolutional network (LSTM-FCN) architecture to process the multivariate time-series data, significantly enhancing landslide detection performance. The experimental results demonstrated that the fused model outperformed models that used only spectral data, achieving higher accuracy and reducing commission and omission errors. Furthermore, the generalization capability of the model was validated on an independent test site, showcasing its potential applicability in diverse geographical contexts. Integrating diverse data sources is essential for accurate landslide detection and can offer valuable insights to facilitate disaster management and mitigation efforts.
ISSN:1947-5705
1947-5713