Abstract: This paper analyzes the defect that the feature points are easily misjudged as ground points in the process of Topcon LiDAR point cloud data processing by progressive TIN densification algorithm (PTD), proposes two kinds of improved methods. The first method is to use the local slope fitting method to improve the PDT algorithm, and sort the point cloud data according to the difference between the elevation value and the fitting elevation value solved by the fitting slope method from small to large, the point which is more greater possibility for the ground point is determined firstly, so as to obtain a more precise tin. The second method is to use thin plate spline(TPS) interpolation to improve the PTD algorithm, change the judgment parameter of candidate point in PTD to the threshold value of bending energy in TPS, so as to reduce misjudgment. The results show that considering the influence of the first error and the second error, the two improved algorithms are better than the traditional PTD algorithm in most terrain features, and have better filtering effect on low vegetation, bridges, slopes and other special objects.

Key words: point cloud, PTD, slope fitting, TPS interpolation, algorithm improvement