An Improved Progressive Tin Densification Algorithm for Lidar Data Filtering Based on Segmentation and Terrain-Adaptive Parameters

摘要

Filtering is one important step in the post-processing of the LiDAR point-cloud data. The progressive triangulated irregular network (TIN) densification (PTD) filtering is widely recognized. However, the PTD sometimes filters the ground as non-ground points in rugged terrain and it is sensitive to threshold parameters that are manually set. To mitigate both shortcomings, we developed a new algorithm using the techniques of the segmentation and terrain-adaptive threshold parameters. A benchmark dataset provided by ISPRS was employed to compare the performance of our and three widely-recognized LiDAR filtering algorithms. The total error (5.54%) and Type I error (4.37%) produced by our algorithm was the smallest in separating the ground and non-ground points. The Type II error was 15.50%. The DEM derived from the filtered ground points consisted of characteristics for rugged terrain. Thus, the developed algorithm was valid and effective.