Research On Point Cloud Data Processing And Modeling Of 3D Laser Scanning

3D laser scanning is a relatively new measurement technology that has revolutionized the fields of big data and virtual reality.It can intelligently,non-contact and efficiently collect point cloud data on the surface of target entities.The constant promotion of the field enriches the way people access data.In order to further analyze the features and shapes of the scanned object surface,it is necessary to convert the irregularly distributed and high-density point data into surface model information to restore the contour features of the object as realistically as possible.Building model information from point cloud data relies on continuous improvement of data processing technology.In this paper,Chongqing Jiaotong University Matilda story building and the statue of point cloud data and data subjects bunny,multi-view clouds registration,theoretical approaches at different scales of noise removal and Modeling methods for different research objects were based on three-dimensional laser scanning system on the principle of Exploring research,the main research contents are as follows:(1)For noise data of different scales,the filtering module in PCL is used to remove large-scale noise by radius filtering and statistical filtering,while the small-scale noise is processed by bilateral filtering and bilateral filtering of optimization weight factors.The bilateral filtering algorithm of the optimization weight factor preserves the local detail features while achieving the desired smoothing effect.(2)Using the open source point cloud library PCL to achieve multi-view cloud registration,the SAC-IA initial pose estimation method based on curvature feature and FPFH operator is applied to the ICP and NDT algorithms respectively and compared with the traditional algorithm.The combination of the planar target and the ICP algorithm applied to the building registration method has effectively completed the registration work.(3)The model and the sculpture entity collected in the field were reconstructed by the contour feature extraction modeling method and the NUBRS surface reconstruction method respectively.The accuracy and applicability of the two reconstruction models were analyzed.Both modeling methods were obtained good result.