测绘通报 ›› 2023, Vol. 0 ›› Issue (8): 45-50.doi: 10.13474/j.cnki.11-2246.2023.0230

• 学术研究 • 上一篇    下一篇

无人机倾斜摄影测量技术支持下的裸露边坡表面变化识别

王宇昊1, 李登华2,3, 丁勇1   

  1. 1. 南京理工大学理学院, 江苏 南京 210094;
    2. 南京水利科学研究院, 江苏 南京 210029;
    3. 水利部水库大坝安全重点实验室, 江苏 南京 210029
  • 收稿日期:2022-11-28 发布日期:2023-09-01
  • 通讯作者: 李登华。E-mail:dhli@nhri.cn
  • 作者简介:王宇昊(1996-),男,硕士生,研究方向为结构健康监测。E-mail:1837594496@qq.com
  • 基金资助:
    国家重点研发计划(2022YFC3005502);国家自然科学基金(51979174);国家自然科学基金联合基金(U2040221)

Surface change identification of exposed slope based on UAV inclined photogrammetry

WANG Yuhao1, LI Denghua2,3, DING Yong1   

  1. 1. School of Science, Nanjing University of Science and Technology, Nanjing 210094, China;
    2. Nanjing Hydraulic Research Institute, Nanjing 210029, China;
    3. Key Laboratory of Reservoir Dam Safety, Ministry of Water Resources, Nanjing 210029, China
  • Received:2022-11-28 Published:2023-09-01

摘要: 为克服传统人工边坡调查方法效率低、风险高、难度大等缺陷,针对高陡裸露边坡,本文提出了基于无人机倾斜摄影的边坡三维重建和灾害识别分类方法,即利用无人机多视角序列影像重构裸露边坡三维实景模型,并将不同时期的边坡三维模型统一在相同的坐标系内。经试验验证,重建模型精度优于2 cm,基于点云与点云比较算法的三维点云数据变化检测算法,能够分析两个时期点云模型的细微差异,通过在三维实景模型中进行标记,结合PointNet++分类神经网络算法自制点云数据集,成功地实现对标记区域的识别与分类,从而实现边坡滑坡、坍塌、落石等灾害场景的智能化识别。

关键词: 无人机倾斜摄影, 边坡灾害识别, 三维重建, 变化监测

Abstract: In order to overcome the defects of traditional manual slope investigation methods such as low efficiency, high risk and difficulty, the 3D reconstruction of slopes and disaster identification and classification method of UAV tilt photography is adopted for high and steep exposed slopes in this paper. The UAV multi-view sequence images are used to reconstruct the 3D realistic model of exposed slopes and unify the 3D models of slopes in different periods in the same coordinate system. The accuracy of the reconstructed model is better than 2 cm after experimental verification. The 3D point cloud data change detection algorithm based on point cloud and point cloud comparison algorithm analyzes the subtle differences between the point cloud models of two different periods and marks them in the 3D real-view model, and combines the PointNet++ classification neural network algorithm with the homemade point cloud data set to successfully realize the identification as well as classification of the marked areas, so as to realize the identification of slope landslides, The intelligent recognition of disaster scenes such as landslide, collapse and rockfall is successfully realized.

Key words: UAV tilt photography, slope hazard identification, 3D reconstruction, change monitoring

中图分类号: