Research On The Key Technology Of DEM Extraction Based On DSM Point Clouds

With the rapid development of photogrammetric technology, the application of airborne LiDAR technology and image matching technology in the field of the geographic national conditions monitoring and the digital 3-D modeling becomes more and more widely. This paper aims at the point cloud data filtering and thinning two aspects.The raw point cloud data is a set of rambling points with some noises, it is necessary to pre-process in a proper way. This paper use grids to segment the raw point cloud data, and the grids are numbered for data management; the noises can be removed by the method mathematical morphology opening and closing on the basis of grids.The point cloud data filtering algorithm based on multi-resolution is described in this paper. The purpose of point cloud data filtering is the classification of ground points and object points, the lowest point of each grids which is selected and marked for the initial ground control points in a certain area; TPS is used for find a smooth surface iteratively, at each level, the control points are filtered according to their residuals based on standard deviation within the local neighborhood. Finally, the object points are filtered based on region growing and TIN; Experiment shows that the method can keep the total error and filter the objects. Unfortunately, the objects in some high complexity of the terrain are misclassified.In order to get the high precision terrain, the massive point data will be used. The redundant data within the point data not only useless for expression of terrain but also takes up a lot of storage space, it is important to remove it. A novel thinning method of point cloud data considering limitation of terrain features is proposed. Firstly, the boundary of the raw point cloud is marked. Secondly the grids for raw point cloud are created to find out the extreme points of each grids, the points of boundary and extreme points is used for seed points to build the triangulation irregular network. Thirdly, the TIN will be reinforced by terrain features points which find out in non-seed points using a specific method step by step. Finally, it is necessary to remove the planar redundant points of the above process by using a planarity testing method of adjacent triangle plane. Thus, the final result will be figure out. The method using the simulated data and the real point cloud data of different terrain to test, the results show that the method can keep the RMSE of results DEM under the standard and thin the raw point cloud at the same time, the quantitative analysis shows that the rate of thin can keep below 35% of raw data. At the same time, in order to improve the practicability of the algorithm, this paper through a large number of experiments to provide the optimal parameters of algorithm.A distributed processing system based on socket is put forward and finished, the system using local area network to management the computer. The purpose of the system is to reduce the manual intervention and improve operational performance by appoint computer for datasets batch processing.