Ground point extraction using self-adaptive-grid and point to surface comparison

Abstract

Extracting accurate ground points by using laser point clouds at the complex terrain of the mining area and the dense vegetation environment is a challenging problem. To address this problem, this paper proposes a combined filtering method based on the Self-Adaptive-Grid method (SAG) and Point to Surface Comparison (PSC). First, the adaptive grid is used to extract initial ground seed points, and then, based on the seed points, a Triangulated Irregular Network (TIN) is constructed, and the distance from the reference surface and TIN is computed. These steps are repeated, In each loop, the non-ground points that are significantly far away from the reference surface will be culled, until the ground point is close to the real terrain, where the initial grid size and the distance from the point to the reference surface are gradually reduced according to the gradient. The proposed algorithm overcomes the difficulty of extracting bare ground due to terrain fluctuations. The ground points are effectively separated from the non-ground points in the laser lidar scan data, and the extracted high-precision ground points can be used as important basic data in the terrain modeling of the mining area, surface subsidence monitoring, and research on the laws of surface subsidence, which has great value in popularization and application.