地理空间信息2026，Vol.24Issue(5)：21-26,31,7.DOI:10.3969/j.issn.1672-4623.2026.05.005
改进U-Net模型在滑坡区域识别中的应用
Application of Improved U-Net Model in the Landslide Area Identification:A Case Study of Xiji County,Ningxia Hui Autonomous Region,China
摘要
Abstract
Loess hilly regions are characterized by complex topography and frequent landslide disasters.We proposed an automated landslide identification method based on improved U-Net model.Compared to the original model,the improved U-Net model achieved an approximately 4.0%increase in overall accuracy and an approximately 8.9%increase in landslide classification accuracy.Focusing on Xiji County,Ningxia Hui Autonomous Region,China,we used the improved U-Net model to identify the landslide areas in high-resolution optical remote sensing images.A total of 7 553 landslide patches were extracted,classified,and the spatial distribution patterns and causes of landslide areas were analyzed.We evaluated the recognition results by combining multi-source datasets with manual visual interpretation.The results demonstrate that the improved U-Net model achieves notable improvements in both identification accuracy and generalization capability,particularly under complex terrain conditions.This method provides a new technical approach for rapid landslide identification in loess hilly regions,offering practical significance for geological disaster prevention and mitigation.关键词
U-Net/滑坡/深度学习/ResNet50/可靠性验证Key words
U-Net/landslide/deep learning/ResNet50/reliability validation分类
天文与地球科学引用本文复制引用
马宁远,杨凯,李吉龙..改进U-Net模型在滑坡区域识别中的应用[J].地理空间信息,2026,24(5):21-26,31,7.基金项目
国家自然科学基金资助项目(42201462) （42201462）
宁夏自然科学基金优秀青年项目(2023AAC05023) （2023AAC05023）
宁夏回族自治区青年科技托举人才培养项目(宁科协发组字[2024]6号) （宁科协发组字[2024]6号）
国家级大学生创新创业训练计划项目(G202410749030). （G202410749030）
