A New Simplification Method Based On Terrain Complexity For LiDAR Point Cloud

LiDAR is a kind of surveying technology which integrates GPS, IMU and Scanning Laser Ranging System working together. As an emerging technology, LiDAR has many advantages, such as high degree of automation, short production cycle, high precision, little impact by the external environmental, and so on. But LiDAR point cloud data volume is great. When using a mass of LiDAR point cloud to build DEM in different scales, it is no significant effects to improve the DEM precision but the huge data will lead to the rapid decline in the speed of data processing. So it is necessary to thin the LiDAR point cloud.When thinning the LiDAR point cloud, the distribution and quality of simplified points is affected directly by the criteria of which point should be reserved. The criteria usually is set based on topography. At present, most of the LiDAR point cloud thin methods are based on the single topography indexes. But the single topography indexes cannot describe and express terrain features comprehensively. So it is necessary to build a terrain complexity index.According to the multi-factor analysis theory, four basic terrain indices were selected: slope, total curvature, terrain relief, terrain roughness to build the terrain complexity index. Based on the principal component analysis method, a terrain complexity construction model has been built. Empirical formula of the terrain complexity index C and four basic terrain indices is presented. And then the correspondence between the terrain complexity index C and topographic features is obtained by experiments, and then the correspondence was verified by experiments. The experimental results show that the terrain complexity index C can describe topographic features effectively and the correspondence between C and topographic features is reasonable and feasible.Based on the terrain complexity index C, a new point sampling rule and a new LiDAR point cloud thin method TCthin have been put forward. The index C would be set as a threshold to thin point cloud. And the new method has been implemented by using the algorithm principle of bounding box. The TCthin method idea is:(1) calculate the index C value based on low-resolution DEM; (2) according to generalization scale, LiDAR point cloud is partitioned into grids which associated with the index C value; (3) according to the index C value, reserve the points which meet conditions. When 0<=C<0.5, Z-Mean point of each grid is reserved; when 0.5<=C<1.5, Z-Max point and Z-Min point of each grid are reserved; when C>=1.5, Z-Mean point, Z-Max point and Z-Min point of each grid are reserved.The new LiDAR point cloud thin method is verified and evaluated by experiments. Eight experimental areas have been selected. The Thin Points method of Software TerraScan and the Lasthin method of Software Lastools are as the comparison methods of the TCthin method. Based on three generalization scale levels, eight sets of LiDAR point cloud datum have been simplified by three thin methods. And then calculate the simplification rate of each simplified dataset and evaluate the thin extent of each dataset according to simplification rate; generate DEM using each thinned dataset by inverse distance weight interpolation method, calculate RMSE of the DEM and evaluate the quality of each thinned dataset. Experimental results show that the new method TCthin can simplify LiDAR point cloud effectively; at the same simplified level, the quality of each thinned dataset by the new method TCthin is superior to the other two methods on a whole; at three simplified levels, the new method TCthin has good applicability in areas with different topographic features.In conclusion, topographic features can be described and expressed comprehensively by the terrain complexity index C. LiDAR point cloud could be simplified effectively by the new LiDAR point cloud thin method TCthin. And the quality of the thinned point cloud could be improved. At three simplified levels, the new method TCthin has good applicability in areas with different topographic features. The study has an important application value in topographic feature expression, thinning LiDAR point cloud, building high-precision DEM and so on.