Research On Point Cloud Of Three-dimensinal Laserscanner Pre-processing Algorithm

Along with 3D measure and computer-related technical development, people can collect datas to obtain the surface information of realistic object by various means, and carry on analysis, handle, process and apply.The point cloud data that is obtained by the scanning got extensive research and application,and a 3D data geometrical process this new research realm is conceived.Pre-processing of point cloud data on the basis of various analysis to the 3D laser scanning technical present condition is select in this paper.Three kinds of bran-new methods for pre-processing of point cloud data are put forward to noise reducing,simplification and registration respectively.These arithmetics are carried out in experiment platform,and proved viable and have efficiently high accuracy.First of all,Savitzky-Golay filter arithmetic based on project to the planar plane is raised.This arithmetic carries on a classification to the point cloud data,different methods are used to point cloud data of different types.Precision of process on the point cloud data improved,and visual evaluation result is given.For simplify,particle-emulation combined with random-sampling simplification arithmetic is put forward.This arithmetic combined accuracy advantage of particle-emulation and the speed advantage of random-sampling.Not only simplify the point cloud data effectively,and well kept the main characteristic in the model at the same time, obtained speed and effect of compromised.For registration,improved ICP arithmetic based on curvature is raised, pre-registration arithmetic of equal-curvature eigen point is adopted combine with an ICP arithmetic precision to align,and have a quite good performance under the sistuation whether the whole model or part model registration. The arithmetics mentioned above,whether respective application or comprehensive application,are all in successful realization.Validity of these arithmetics this paper raised is proved.