基于三维点云的大坝缺陷体积测量方法

Volume measurement of dam defects based on 3D point cloud

摘要:

大坝在运行过程中不可避免的会出现混凝土呈块状剥落的情况。因此，精确的剥落缺陷体积测量对于其结构修缮具有重要意义，而缺陷的形状往往是不规则的，难以通过简单的几何运算求出。为实现非接触精准测量，提出一种将点云平面拟合、滤波和三角化相结合的缺陷体积测量方法。利用单目相机拍摄待测结构物，对其进行多视图三维重建，获取点云数据，分割出缺陷点云并对其进行体素降采样，采用Delaunay三角剖分法计算缺陷体积。为提高测量精度，改良传统RANSAC平面拟合算法，在其中融入统计离群值移除，实现缺陷区域的精准分离。经过多次实验证明，该方法在测量缺陷时，无论形状是否规则，都能较为精准地测量出其真实体积。将测量结果与仅用RANSAC算法分割相比，测量精度提高70.32%。该方法大幅度提升了缺陷的测量精度。

Abstract:

During the operation of dams, the occurrence of concrete spalling is inevitable. Therefore, accurate volume measurement of spalling defects is crucial for effective structural rehabilitation. The irregular shapes of these defects often make simple geometric calculations inadequate. To achieve non-contact and precise measurement, this paper proposes a method that combines point cloud plane fitting, filtering, and triangulation for defect volume measurement. A monocular camera is used to capture images of the structure under examination, and multi-view 3D reconstruction is employed to obtain point cloud data. The defect point cloud is segmented and downsampled using voxel-based methods, and the Delaunay triangulation algorithm is applied to calculate the defect volume. To enhance measurement accuracy, the traditional RANSAC plane fitting algorithm is improved by incorporating statistical outlier removal, enabling precise separation of the defect region. Multiple experiments have demonstrated that this method accurately measures the true volume of defects, regardless of their shape. Compared to segmentation using the RANSAC algorithm alone, the measurement accuracy is improved by 70.32%. This approach significantly enhances the precision of defect volume measurement.