Study Of 3D Point Cloud Data Adaptive Simplification

Point cloud data,as a kind of 3D information representing object shape,its application covers various fields in our daily work and life.With the enlargement of its application field,the study of the point cloud data increasingly attracts the attention of scholars from all walks of life,especially the point cloud pretreatment,including denoising,segmentation,simplification,etc.It has a great influence on the subsequent point clouds reconstruction work.This thesis mainly aims at the point cloud simplification for in-depth research.Usually original point cloud data is quite large.The necessity of research on point cloud simplification lies in that it is not reasonable to use so many data directly.Though some of the data can reflect the object shape,the original data contains noise and redundant information,which leads to the computer inefficient in running,storage,and operation,and then,affects the later work of point cloud post-processing,etc.It needs to optimize the morphology information,extract the useful information,remove the redundancy and noise points.On the premise of ensuring accuracy,it realizes the goal of taking the smallest space to show the most effective information.Based on several classic point cloud simplification algorithms,this thesis puts forward the point cloud of feature extraction adaptive simplification algorithm: first,it divides space to the original point cloud,builds k neighborhood of the point,sets up the feature parameters,analyzes feature,identifies the different parts of the information and data of the original point cloud,and divide into planar data and point cloud data with curvature.For the planar data,it detects and extracts the boundary,simplifies the rest;For the data with curvature,it extracts the feature and then simplifies the rest in varying degrees according to the different curvature.The 3D scanning devices Kinect and GO!SCAN have been used to capture 3D point cloud data on real scene and the models.This article uses the PCD file format.Under the Microsoft visual studio 2010 platform,C++ programming is used to realize the point cloud feature extraction and adaptive simplification.Through comparing with the experimental results of other algorithms,it proves that the data processing with the proposed algorithm can be better to show the object shape.Finally,it has carried on the comprehensive evaluation on three aspects including streamline speed,reduction and precision.The plane fitting and sphere fitting combination method was used for precision verification.Some actual related experiments have been carried out to evaluate the precision of the proposed method.