Abstract: As an advanced remote sensing technology, airborne LiDAR is one of the important means to obtain high-precision digital elevation models (DEM) of vegetation coverage area. Different terrain slope conditions and point cloud density have an important impact on the accuracy of DEM products. Based on airborne LiDAR point data from a city in Liaoning province, this paper selects the point cloud data with different terrain slopes from 0° to 50° and adopts three different point cloud thinning methods:random, equidistant and curvature. The original point cloud is thinned and simplified according to the thinning multiple of five different gradients with the point cloud retention rate of 80%,60%,40%,20% and 10%,and the corresponding DEM is generated and its accuracy is evaluated. Based on this, the influence of terrain gradient, point cloud thinning method and thinning multiple on DEM accuracy is studied. The results show that there is a negative correlation between DEM accuracy and terrain slope, that is the RMSE increases with the increase of terrain slope.Compared with the other two methods,the RMSE of curvature based thinning method is smaller when the terrain slope is greater than 30° and has obvious advantages.When the point cloud retention rate is less than 40%, the RMSE of DEM will increase rapidly.This research has certain guidance and reference significance for large-scale DEM production using airborne LiDAR.

Key words: airborne LiDAR, digital elevation model, point cloud density, accuracy, point cloud thinning