Abstract
Extracting ground points from LiDAR point clouds is essential for constructing digital terrain models (DTMs). This article presents a filtering method with strip curve fitting, based on the linear scanning characteristics of airborne LiDAR and the concept of limits. The entire point cloud is divided into multiple strips, each projected onto the YZ plane to simplify the problem from three dimensions to two. Each strip contains several linearly scanned point clouds, and the ground points from these scans form a curve on the YZ plane. To obtain the curve, an initial fitting point is selected from the strip, which is then segmented based on slope changes and curvature. Within each segment, the difference between the left and right slopes of each point is calculated, and various thresholds are applied to screen the initial fitting points. By iteratively repeating this screening process, points that do not meet the criteria are progressively eliminated, ultimately yielding stable fitting points. These final fitting points are used for curve fitting to generate the target curve. Data points in proximity to the curve are identified as ground points and retained. The residual is defined as the difference in the Z-coordinate between each data point and the target curve. The retained point cloud is determined by combining Tukey’s fence with a residual threshold. Reference data provided by the International Society for Photogrammetry and Remote Sensing (ISPRS) Working Group III/3 were used to quantitatively describe the accuracy of the method. OpenTopography data and Airborne LiDAR scan point cloud data visually show the filtering effect of dense point clouds. Experimental results on different datasets show that the proposed filtering method can effectively remove most nonground points, especially for side-slope point cloud filtering.
Funding Information
  • National Key Research and Development Project of China (2021YFB2600603)
  • Natural Science Foundation of Chongqing, China (CSTB2022NSCQ-MSX1527)

This publication has 29 references indexed in Scilit: