基于机载LiDAR-DEM多特征信息融合增强的滑坡裂缝识别  

Landslide crack identification enhanced by multi-feature fusion based on airborne LiDAR-DEM

作　　者：韩磊 段平[1] 王扶卷 伍红玲[3] 李佳[1] HAN Lei;DUAN Ping;WANG Fujuan;WU Hongling;LI Jia(School of Geographical Sciences,Yunnan Normal University,Kunming 650000,China;Petroleum and Chemical Engineering Quality Supervision Station,Beijing 100018,China;Yunnan Provincial Basic Geographic Information Center,Kunming 650000,China)

机构地区：[1]云南师范大学地理学部,云南昆明650000 [2]石油化工工程质量监督总站,北京100018 [3]云南省基础地理信息中心,云南昆明650000

出　　处：《自然灾害学报》2025年第2期79-88,共10页Journal of Natural Disasters

基　　金：国家自然科学基金项目(42361070);云南省基础研究计划项目(202401AT070125,202301AT070061);兴滇英才支持计划项目(YNWR-QNBJ-2020-048,YNWR-QNBJ-2020-103);云南省基础研究专项重点项目(202201AS070024);云南省高层次科技人才及创新团队选拔专项(202305AS350003)。

摘　　要：数字高程模型(digital elevation model,DEM)可用于表达滑坡裂缝凹陷地形特征及高程变化,基于DEM能够对滑坡裂缝进行识别。然而,单一的DEM地形地貌数据难以精确表达滑坡裂缝复杂的地形特征。文中提出了一种基于机载激光雷达(light detection and ranging,LiDAR)DEM多特征地形地貌数据融合增强的滑坡裂缝识别方法。首先,通过无人机挂载LiDAR传感器采集获取裂缝的点云数据,并构建LiDAR-DEM,基于DEM生成4种多特征地形地貌数据,包括山体阴影、坡度、正开度和天空视域因子。其次,利用考古地形可视化(visualization for archaeological topography,VAT)方法对DEM多特征地形地貌数据进行融合增强处理,得到DEM多特征融合增强图像。然后,基于大津法(Otsu’s method,OTSU)算法进行滑坡裂缝识别。最后,采用连通域分析和数学形态学闭运算对识别出来的滑坡裂缝进行优化,得到完整准确的滑坡裂缝。以云南省禄劝县的烂泥沟滑坡裂缝为研究区,采用精确率、召回率、F1分数指标对识别结果进行精度验证。最终精确率为0.85,召回率为0.90,F1分数为0.87。研究结果表明,文中方法的滑坡裂缝识别效果较好,能够得到完整的滑坡裂缝,可为滑坡裂缝识别提供一定的参考。The digital elevation model(DEM)can be used to express the terrain features and elevation changes of landslide cracks,and landslide cracks can be identified based on the DEM.However,the single DEM terrain data is not sufficient to accurately express the complex terrain features of landslide cracks.This paper proposes a method for landslide crack identification enhanced by multi-feature terrain data fusion based on airborne light detection and ranging(LiDAR)DEM.Firstly,the point cloud data of cracks are collected via drones equipped with LiDAR sensors to construct the LiDAR-DEM,and four types of multi-feature terrain data are generated based on the DEM,including mountain shadows,slope,positive openness,and sky view factor.Secondly,the visualization for archaeological topography(VAT)method is used to enhance the fusion of the multi-feature terrain data from the DEM,resulting in an enhanced multi-feature fusion image of the DEM.Then,the landslide crack identification is performed based on Otsu’s method(OTSU).Finally,connected component analysis and mathematical morphological closing operations are used to optimize the identified landslide cracks,resulting in accurate and complete landslide cracks.The Lannigou landslide cracks in Luquan County,Yunnan Province,were used as an example.The identification results were validated by using precision,recall,and F1-score metrics.The final precision was 0.85,recall was 0.90,and F1-score was 0.87.Experimental results show that the method presented in this paper is effective in landslide crack identification and can provide a reference for the identification of landslide cracks.

关 键 词：机载激光雷达 数字高程模型 多特征图像融合 VAT 滑坡裂缝识别 

分 类 号：P642.22[天文地球—工程地质学] X43[天文地球—地质矿产勘探]