Study Of Reduction Based On Scattering Point Cloud Data

In the surveying and mapping,3D laser scanning is a kind of new techniquewhich scanning all of object high-precision, fast, non-contact, continous in the spaceso now it is a important way to get3D coordinate of object and it is widely used in thehand of protection of cultural relics, the large complex building’s construction quality,deformation monitoring and disaster assessment survey. As the companies at homeand abroad cost lots of money in develop hardware and software, the hardware andsoftware develop in the unprecedented speed.With the rapid development technology,there is a urgent problem that whether manage so much data or use them to do somereal research. Because writer acquisite the data in the outdoor, the point data whichWriter scanned is not necesary in some cases and is very large. Take the big clouddragon stone carving as example,for aciquisite the full of object’s information Writermust scanning the stone carving in different hundreds of places and take hundreds ofHD photograph.you must know there are more than10g data which discribles thebig cloud dragon stone carving’s3D which is only one historic building.Writerdeduce that Writer must acquisite dozens of data for discribling the whole palacemuseum.for the aim of using the point data in some different cases,Writer using lesspoint data to discrible important and complicated feature of the object in the space.Because Writer need find the useful point from the big point data,Writermust build the spatial index for all data.Writer study kinds of spatial index such asKD-tree,R-tree, oc-tree and compare the utility of the three algorithms. Writer needto search K-area to finish the other works of the essay that those are fitting surfaceand calculate normal,so Writer select the KD-tree as the spatial index of index that isused in the essay.Writer read many papers about curvature sampling for selecting thealgorithm that can simplify the cloud and then remain the curvature feature point.Forthe purpose of maintain the real feature of object,writer select the Gauss curvaturealgorithm to simplify the point.Why write select Gauss curvature algorithm. BecauseGauss curvature and Draw the curvature can discrible how anfractunous the object’ssurface.But for finishing the study, there are one thing that is dealing with is speedingup the operation.Writer find that The no-linear calculate is big problem whencalculating the normal vector.For solving the problem,writer select GPU multi-threadcomputing of CUDA framework to speed up the calculating normal vector.This paper chooses Visual Studio2010as development platform and is based on the CLR (common language runtime) environment, then uses C++developmentlanguage with PostgreSQL database and develops the database management modulein TLiDAR_DPS system by the laboratory。In the end writer achieve the fellows asfunction: the original point cloud data of input and output、the index build of theoriginal point cloud、calculating the normal vector、curvature reduction、Gridreduction、Random reduction. For the purpose of examining the accuracy of result, in thesoftware system writer designs the interactive operation easily and conveniently.Inthe end the result turned out that the study content of the paper is feasible and useful.