Research On Surface Surface Reconstruction Based On 3D Laser Scanning Data

The establishment of virtual geographical environment involves in multiple technologies. The development of new geography information acquisition technology, provides a good data support platform for virtual geographic environment. In the moment of data collection technology, the 3D laser scanning technology has its unique advantages which provide a means that any kind of technology can not be compared to the real world of digital. In order to reconstruct the 3D model of the earth’s surface, the first thing to do is to carry out a full range of surface scanning. Due to the impact of the site environment, observation instrument and surface shape, it can not get all surface point-cloud data through a single scan, so, obtain the surface from different angles of the point-cloud data by 3D laser scanner at first. And then splicing the different angles of the point-cloud data, transformed into the same coordinate system. In the process of coordinate conversion, the quality of algorithm determines the data precision after conversion. The author has put forward to the coordinate conversion method which based on a plurality of public control point. It has made a good transition effects. Finally,started the reconstruction of earth curved surface.During the process of reconstruction, since the acquisition of massive point clouds, in order to improve the efficiency of data processing, put forward to the method of using BP neural network for surface reconstruction, which have achieved good experimental effect. The main work and results are as follows:(1) the point-cloud coordinate transformationUsing two kinds of coordinate conversion algorithm of coordinate of point- cloud data conversion, obtained by comparing the experimental results analysis, using the coordinates of a plurality of control points based on the conversion accuracy were significantly higher, also ensure the data precision of later surface reconstruction.(2) the point-cloud data compressionComparing the traditional octree division point-cloud data compression method with weighted hierarchical point-cloud compression method, The results obtained show that the weighted hierarchical point-cloud compression method which can better retaining the surface details, more suitable ground surface reconstruction.(3) analysis of B spline least squares on surface reconstructionIt has achieved B spline least squares approximation algorithm in MATLAB, and the experimental results show that the fitting precision of this method is very high, but we are getting mass point-cloud data, and this method need to calculate for each point by least square, a large quantity of operation and time. Such as use of fewer points to the fitted surface continuity is poor, and will lost a lot of ground surface detail information of surface reconstruction, which contrary to the original intention of the use of 3D laser scanning technology.(4) the BP neural network method on surface reconstruction methodIt has achieved the surface reconstruction based on BP neural network algorithm in MATLAB, and achieved very good results, point cloud data quantity is big, the precision is high, the amount of calculation is almost no effect, therefore written procedures can also carry on the superposition of two data model comparison on the same surface, to obtain any changes in the surface.