基于贴近摄影测量的崩岗侵蚀监测技术

在崩岗侵蚀研究中,为了高效精准的对高陡崩壁土体侵蚀沉积运移进行准确监测,运用无人机贴近摄影测量技术对崩岗研究区进行数字影像采集,通过运动恢复结构-多视点匹配(Structure From Motion-Multi View Stereo,SFM-MVS)技术,生成点云数据及研究区数字表面模型(Digital Surface Model,DSM),利用ArcMap进行叠加分析监测周期内研究区侵蚀沉积变化。从定位精度、测量精度、重现性分析3个方面对贴近摄影测量技术进行误差及可行性分析。最终验证结果贴近摄影测量技术总平均重投影误差为0.19 mm,侵蚀沉积量总平均绝对误差为0.006 m~3,较传统倾斜摄影测量技术误差降低了了45.45%。高程精度较倾斜摄影测量技术总体提升了162.5%。重复点云数据高程误差的平均值仅为0.36 mm,识别图像及控制点误差均达到毫米级,因此利用无人机贴近摄影测量技术精度满足崩岗高陡崩壁监测需求,该技术可提取崩岗研究区侵蚀地貌特征信息,是较为高效精准的研究侵蚀沉积过程的监测技术。

Monitoring technology for collapse erosion based on the nap of the object photograph of UAV

Abstract: Collapsing gully is one of the most serious ecological and environmental damage in southern China. The dynamic change of wall collapse is also difficult to monitor in the high-risk and complex terrain. Obviously, a more concise and accurate method is necessary to monitor the dynamic collapse, and further to clarify the collapsing gully erosion. Unmanned Aerial Vehicle (UAV) nap of the object photograph is brand new photogrammetry for the needs of fine measurement. Point cloud data was generated to extract different parameters of collapse terrain. The topographic features of the avalanche wall were extracted, and then the obtained data was used to analyze the spatial characteristics and soil migration changes under the photogrammetry. Collapsing gully research area was located in the AnXi County, Fujian Province of South China. The specific procedure was as follows. 1) The control points were arranged in the study area, where the coordinates of each control point were measured using Trimble R4 GNSS GPS, Trimble R2 GPS, and (1+1) RTK network mode. 2) The image data was acquired by the nap of photogrammetry technology. PIX4DMapper was utilized to evaluate the positioning accuracy of control point data obtained by RTK. 3) The Digital Surface Model (DSM) data in the study area was processed by ArcMap, whereas, the digital surface model of differences (DOD) values was obtained for measured accuracy. The DOD was equivalent to the change of soil erosion and deposition volume, accurately reflecting the volume change of soil dynamic migration. The elevation change was also measured in the research area during the monitoring period. Through CloudCompare, Scale-invariant feature transform-Clustering Views for Multi-View Stereo (SFM-MVS) technology was used to analyze the reproducibility of point cloud data. A Canon 5D Mark III SLR camera was applied to the surround fixed focus-range image to collect data, and simultaneously to measure the target wall collapse as the reference data. Three groups of control plots were arranged according to the lowest level (LODmin) in the study area. The error sources and feasibility close to photogrammetry were analyzed from the accuracy of positioning and measurement, as well as the point cloud reproducibility. 4) The distribution characteristics of landslide erosion deposition were evaluated during the monitoring period, combined with the rainfall data collected by an RG3-M rainfall recorder. The results showed that: 1) The average resolution of the image close to the photogrammetry technology was 4.1 mm in terms of positioning. The average projection error was controlled at about 0.19 mm in terms of control point calibration. 2) The total mean absolute error of DOD in measurement was 0.006 m3, 45.45% lower than that of traditional oblique photogrammetry. 3) The overall improvement was 162.5% in elevation accuracy. 4) In reproducibility, the average value of elevation error was only 0.36 mm for the repeated point cloud data, indicating high stability. 5) The dynamic change of erosion settlement in the collapsed wall was monitored in the study area during the study period. The total rainfall was 564.6 mm in the study area, while the soil loss of the collapsed wall was 4.758 m3. The UAV Nap of the object photograph can be expected to serve as efficient and accurate monitoring for the erosion sediment transport in south collapsing gully areas of China.