Analysis of the Impact of Optimized Airborne LiDAR Point Density on DEM Accuracy under Different Terrain Conditions

Airborne LiDAR point cloud data, as a commonly used spatial data, is also the main source for obtaining high-quality Digital Elevation Model (DEM) at present. In order to study the relationship between point cloud data and interpolation algorithms under different sampling densities on the accuracy of DEM construction, and to verify that the test area can form high-quality DEM topographic maps in 1:500 and 1:1000 large-scale topographic maps to meet the mapping standards, this paper uses airborne point cloud data under flat and gully terrain conditions as the experimental objects. Firstly, the KD tree is constructed to find the effective point cloud according to the distance threshold, in order to further enhance the filtering effect, the bilateral filtering factor is used to determine the noise points to achieve multi-scale denoising, and then the fabric filtering algorithm is used to find the seed points to achieve the accurate separation of the ground points from the non-ground points; secondly, the ground point data are randomly sampled into different densities, and the ground point data are randomly sampled into different densities by establishing a spatial interpolation algorithm (Kriging (Ordinary Kriging, OK), Radial Basis Function (RBF), and Inverse Distance Weighting (IDW) to generate DEMs at different densities, and use the root-mean-square error for accuracy assessment. The results show that: 1. The accuracy of the DEM generated by the interpolation algorithm all decreases with the reduction of the point cloud density, and the accuracy of different interpolation algorithms is obviously different, when the sampling rate reaches 50% OK and IDW accuracy is optimal, the topographic features are clearly characterised and the RBF accuracy is the lowest; 2. With the enhancement of the density of point cloud on the ground, the level of accuracy of the DEM tends to be gradually stabilised. When the sampling density reaches below 30%, the DEM accuracy decreases linearly.