基于三维地形扫描的坡沟系统侵蚀产沙监测方法对比  

Comparison of Erosion Monitoring Algorithms Used for Slope-Gully System Based on 3D Laser Scanning

作　　者：李朋飞 王锦博 张晓晨 胡晋飞 刘立峰[2] 高健健 党恬敏 Li Pengfei;Wang Jinbo;Zhang Xiaochen;Hu Jinfei;Liu Lifeng;Gao Jianjian;Dang Tianmin(College of Geomatics,Xi′an University of Science and Technology,Xi′an 710054,China;Suide Test Station of Soil and Water Conservation,Yellow River Conservancy Committee of Ministry of Water Resources,Yulin,Shaanxi 719000,China;Yellow River Basin Monitoring Center of Water-Soil Conservation and Eco-Environment,Xi′an 710021,China)

机构地区：[1]西安科技大学测绘科学与技术学院,西安710054 [2]黄河水利委员会绥德水土保持科学试验站,陕西榆林719000 [3]黄河流域水土保持生态环境监测中心,西安710021

出　　处：《水土保持研究》2024年第4期75-85,共11页Research of Soil and Water Conservation

基　　金：国家重点研发计划政府间国际科技创新合作重点专项项目(2022YFE0119200);国家自然科学基金(41977059;U2243211;42207407);水利部重大科技项目(SKS-2022092);陕西省自然科学基金(2022JQ-259);陕西省教育厅基金(22JK0463)。

摘　　要：[目的]阐明不同算法在坡面侵蚀监测中的精度和适用性,进而为土壤侵蚀过程监测算法的选择和构建提供参考。[方法]于黄土丘陵沟壑区典型流域同一自然坡面建立5个小区进行径流冲刷试验,以TLS三维点云数据为基础,通过DEM of difference(DoD)、Cloud to Cloud(C2C)、Cloud to Mesh/Model(C2M)和Multiscale Model to Model Cloud Comparison(M3C2)等方法计算侵蚀产沙量,并分析了不同算法对于侵蚀产沙的监测差异。[结果]不确定性分析结果表明:M3C2平均不确定性最小,C2C,C2M次之,DoD最大。产沙结果表明:大流量(85,70,55 L/min)下,4种算法单场次和累计场次产沙量与实测产沙量之间有显著的线性关系(R 2>0.62,p<0.05),M3C2表现最优;小流量(40,25 L/min)下,单场计算产沙量与实测产沙量之间的线性关系不显著但累计产沙量与实测产沙量之间有显著的线性关系(R 2>0.91,p<0.05),DoD表现最优。侵蚀沉积空间分布特征表明:C2C,M3C2和DoD均能反映梁峁坡和沟谷坡侵蚀发展经历的两个阶段(快速发育和稳定发育),其中M3C2能够检测到细微的地形变化,但在TLS扫描盲区,M3C2由于在法线方向上未找到对应点会出现“空洞”。[结论]M3C2算法更适合监测复杂三维地形,但在扫描盲区仍会失效,未来应改进算法,有助于应对更加复杂和剧烈的地形变化。[Objective]The aim of this study is to elucidate the accuracy and applicability of different algorithms for detecting slope erosion,which can provide a reference for the selection and development of detection methods for soil erosion processes.[Methods]Five plots established on a natural slope in the loess hilly and gully region to conduct runoff scouring experiments.Using TLS data,we calculated erosion and sediment yield by using various methods such as Digital Elevation Model(DEM)of difference(DoD),Cloud to Cloud(C2C),Cloud to Mesh/Model(C2M),and Multiscale Model to Model Cloud Comparison(M3C2),and compared their results.[Results]The results of the uncertainty analysis showed that M3C2 produced the smallest average uncertainty,followed by C2C and C2M,while DoD yielded the largest uncertainty.The sediment yield calculated by the detection algorithms demonstrated that under high flow rates(85,70 and 55 L/min),there was a significant linear relationship between consecutive sediment yield and cumulative sediment yield derived by the four algorithms and the corresponding measured sediment yield(R 2>0.62,p<0.05),and M3C2 performed best;under low flow rates(40 and 25 L/min),no significant linear relationships were found between consecutive sediment yield and measured sediment yield,but significant linear relationships were found between cumulative sediment yield and measured sediment yield(R 2>0.91,p<0.05),and DoD performed best.The spatial distribution of erosion and deposition indicated that C2C,M3C2,and DoD could reflect two stages of erosion evolution on hillslopes and gully slopes(rapid development stage and stable stage),with M3C2 being able to detect subtle topographic changes.However,M3C2 results were subject to‘voids’in the blind area of TLS scanning due to not finding corresponding points in the normal direction.[Conclusion]The M3C2 algorithm is more suitable for detecting complex terrain,but it will still fail in the blind spot of scanning,and the algorithm should be improved in the future to help cope

关 键 词：三维激光扫描 地形变化监测 误差分析 黄土高原 

分 类 号：S157[农业科学—土壤学] P237[农业科学—农业基础科学]