Objective To improve the accuracy of ground point extraction from unmanned aerial vehicle structure-from-motion (UAV-SfM) point clouds, and to construct high-precision digital elevation model (DEM) for complex terrain in the hilly and gully region of the Loess Plateau. Methods At large spatial scales, this study utilized a spatial distribution feature-based ground extraction (SDFGE) method that considered near-ground point clouds using UAV-SfM data. It further investigated the effects of plot scale, terrain slope, vegetation coverage, and point cloud density on ground point extraction accuracy. Results 1) The Type I error, Type II error, and total error of ground point extraction using the SDFGE algorithm were 13.68%, 12.92%, and 13.40%, respectively, showing reductions of 1.16%, 6.02%, and 2.79%, respectively, compared with the CSF algorithm. 2) For DEM construction, the SDFGE algorithm had an average error of 0.56 m, 23.29% lower than that of the CSF algorithm (0.73 m). The standard deviation was 1.57 m, 13.26% lower than that of the CSF algorithm (1.81 m). 3) A quantitative evaluation was conducted on the effects of plot scale, terrain slope, vegetation coverage, and point cloud density on ground point extraction. Conclusion The proposed SDFGE algorithm significantly improves the accuracy of ground point extraction in the hilly and gully region of the Loess Plateau by integrating the spatial distribution features of near-ground point clouds, offering a low-cost and efficient solution for obtaining high-precision DEM data for soil and water conservation monitoring.
传统的DEM制作方法主要是人工实地测量,其主要依赖于精确的地面控制点和手动绘制等高线。通过地面测量方法能够勾画出一定精度的地形图,并通过等高线表示地表的高程变化。然而,DEM的制作过程繁琐、费时费力且精度有限,适用于小范围、局部区域的地形测量。随着遥感技术发展,地形数据获取逐渐转向更高效、更精确的远程探测手段。有研究[4]已基于遥感数据制作高精度的DEM,依据所使用的数据源的不同可将常见的DEM制作方法划分为两类。第1类是基于激光雷达(light detection and ranging,LiDAR)数据的DEM制作,基于改进型三角网加密算法(ATIN)、布料模拟滤波算法(CSF)及多尺度曲率分类算法(MCC)等提取地面点并生成DEM[5-7]。邹正等[8]对以上3种算法的点云滤波结果进行对比分析发现,CSF算法总误差最低,性能最佳。在复杂地形条件下进行一项高密度点云滤波测试[9]表明,简单形态学滤波算法(SMRF)在城市地区表现最佳,而多尺度分层滤波算法(MHF)在森林地区表现更为优异。然而,LiDAR技术受制于设备集成复杂、核心部件制造成本高、数据采集与处理要求高等因素[10],在大范围、长周期监测任务中的应用受到一定限制。第2类是基于运动结构恢复算法(SfM)技术生成DEM,基于SfM技术的无人机摄影测量具有成本低、操作灵活的优势,基于SfM技术对采集影像进行处理可获取高精度点云数据(SfM点云),适用于大范围区域的地形数据获取和高分辨率的DEM制作[4,11-12]。在目前针对SfM点云的研究中,CSF算法被认为是进行地面滤波的最精确算法之一[13],并且不断有新的算法被提出,提供更可靠和准确的过滤结果,如ŠTRONER等[11]提出,CANUPO和CSF结合的算法去除地面上的植被部分,得到总误差为0.9%,是非常适用于陡坡植被区的滤波方法。摄影测量技术的精度和可靠性逐渐提升,使其成为替代高成本激光雷达技术的重要选择。然而,目前对于SfM点云的滤波算法研究,多局限于小尺度或简单地形条件的点云过滤,但是对于大尺度或具有复杂地貌的山区地形依然是比较复杂的问题。其次,基于SfM点云植被-地面分离存在过滤不彻底的问题[6,14],导致生成的DEM数据存在较大误差。现有研究[15]多聚焦于单一景观下滤波算法的性能比较,而系统分析地形与环境因素对滤波算法影响的研究较为缺乏。因此,基于SfM点云数据在大尺度区域进行高精度的地面点与非地面点分离研究,并量化各环境因素对滤波算法性能的影响机制具有重要的研究意义。
LIP F, LID, HUJ F, et al. Assessing the ability of airborne LiDAR to monitor soil erosion on the Chinese Loess Plateau[J].Acta Geodaetica Et Cartographica Sinica,2023,52(8):1342-1354.
ZHOUY Y, DAIW, WANGC, et al. Spatial resolution of digital elevation models on fine-scale topographic change detection[J].Mountain Research,2023,41(3):446-458.
LIUY L, LIP F, LID, et al. Comparison of erosion monitoring methods in the pisha sandstone areas of the Chinese Loess Plateau based on UAV-SfM data[J].Journal of Soil and Water Conservation,2024,38(3):91-100.
HUIZ Y, CHENGP G, GUANY L, et al. Review on airborne LiDAR point cloud filtering[J].Laser and Optoelectronics Progress,2018,55(6):7-15.
[11]
ANDERSN, VALENTEJ, MASSELINKR, et al. Comparing filtering techniques for removing vegetation from UAV-based photogrammetric point clouds[J].Drones,2019,3(3):e61.
[12]
ZEYBEKM, ŞANLIOĞLUİ. Point cloud filtering on UAV based point cloud[J].Measurement,2019,133:99-111.
ZOUZ, ZOUJ G, HUH Y. Comparative analysis on different airborne LiDAR point cloud filtering algorithms[J].Journal of Geomatics,2021,46(5):52-56.
[15]
CHENC F, GUOJ J, WUH M, et al. Performance comparison of filtering algorithms for high-density airborne LiDAR point clouds over complex landscapes[J].Remote Sensing,2021,13(14):e2663.
LIUJ C, GUOY J, ZENGJ, et al. Forest point cloud registration using the tree top and the ground-level tree center[J].Transactions of the Chinese Society of Agricultural Engineering,2024,40(15):127-134.
[18]
ŠTRONERM, URBANR, LIDMILAM, et al. Vegetation filtering of a steep rugged terrain: The performance of standard algorithms and a newly proposed workflow on an example of a railway ledge[J].Remote Sensing,2021,13(15):e3050.
[19]
ENWRIGHTN M, KRANENBURGC J, PATTONB A, et al. Developing bare-earth digital elevation models from structure-from-motion data on barrier islands[J].Isprs Journal of Photogrammetry and Remote Sensing,2021,180:269-282.
[20]
SERIFOGLU YILMAZC, YILMAZV, GÜNGÖRO. Investigating the performances of commercial and non-commercial software for ground filtering of UAV-based point clouds[J].International Journal of Remote Sensing,2018,39(15-16):5016-5042.
[21]
LIAOJ H, ZHOUJ X, YANGW T. Comparing LiDAR and SfM digital surface models for three land cover types[J].Open Geosciences,2021,13(1):497-504.
[22]
ZHAOX Q, SUY J, LIW K, et al. A comparison of LiDAR filtering algorithms in vegetated mountain areas[J].Canadian Journal of Remote Sensing,2018,44(4):287-298.
[23]
ZHANGW M, QIJ B, WANP, et al. An easy-to-use airborne LiDAR data filtering method based on cloth simulation[J].Remote Sensing,2016,8(6):e501.
[24]
GAOF, LIUY, SHIP B, et al. Dual-scale point cloud completion network based on high-frequency feature fusion[J].Image and Vision Computing,2023,139:e104818.
ZHANZ Q, HUM Q, MANY Y. Multi-scale region growing point cloud filtering method based on surface fitting[J].Acta Geodaetica Et Cartographica Sinica,2020,49(6):757-766.
ZHANGH C, ZHOUL W, BIANL M. Analysis of drought phenotypic characteristics of poplar seedlings based on TLS point cloud skeleton extraction[J].Transactions of the Chinese Society for Agricultural Machinery,2025,56(3):188-197.
CUIJ R, PUY W, XIAY, et al. Airborne laser point-cloud filtering in complex mountainous terrain utilizing deep global information fusion[J].Laser and Optoelectronics Progress,2024,61(18):228-237.
[31]
EVANSJ S, HUDAKA T. A multiscale curvature algorithm for classifying discrete return LiDAR in forested environments[J].IEEE Transactions on Geoscience and Remote Sensing,2007,45(4):1029-1038.
[32]
ZHAOX Q, GUOQ H, SUY J, et al. Improved progressive TIN densification filtering algorithm for airborne LiDAR data in forested areas[J].Isprs Journal of Photogrammetry and Remote Sensing,2016,117:79-91.
[33]
KRAUSK, PFEIFERN. Determination of terrain models in wooded areas with airborne laser scanner data[J].Isprs Journal of Photogrammetry and Remote Sensing,1998,53(4):193-203.
[34]
CRESPO-PEREMARCHP, TORRALBAJ, CARBONELL-RIVERAJ P, et al. Comparing the generation of DTM in a forest ecosystem using TLS, ALS and UAV-DAP, and different software tools[C].The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences,2020,XLIII- B3-2020:575-582.
MAJ C, GUOZ, XUH, et al. Effect of forest structure complexity on single wood segmentation accuracy: A case study of Tianheng Island[J].Acta Ecologica Sinica,2024,44(11):4770-4781.
TANGF Q, YANGQ. Progress and prospects of multi-source remote sensing monitoring technology for coal mining subsidence in mining areas of the western Loess Plateau[J].Coal Science and Technology,2023,51(12):9-26.
LID, LIP F, MUX M, et al. Accuracy of airborne LiDAR point cloud filtering for areas with complex terrain[J].Research of Soil and Water Conservation,2021,28(4):171-178.
LEIQ J, LIUJ, CAOX Y. Accuracy evaluation of open DEM products based on airborne LiDAR data[J].Geomatics and Information Science of Wuhan University,2025,50(1):153-163.
BEIY X, CHENC F, WANGX, et al. Effects of airborne LiDAR point cloud density and interpolation methods on the accuracy of DEM and surface roughness[J].Journal of Geo-Information Science,2023,25(2):265-276.