Research On Spatial Scattered Points Cloud Denoising, Streamlining And Surface Reconstruction Algorithm

In recent years, with the development of the 3D scanning equipment, the way of getting the data of the object surface becomes easier and easier. The data that we get are called points cloud data. From points cloud data into geometry object, we need to pass through the data denoising, data simplification, surface reconstruction and surface subdivision to get the model that satisfies the users. This technology can be widely used in computer aided geometric design, computer vision, medical image processing and other related fields. This paper is based on the study of the relevant research home and abroad on the algorithm of points cloud processing. It is mainly about the points cloud denoising, simplification and surface reconstruction. The specific work is below:Aiming at the problem of points cloud denoising, after analysing the reason of noise, noise model, common used algorithm of points cloud, a new denoising algorithm based on the traditional bilateral algorithm is proposed. The algorithm absorbs the advantage of bilateral algorithm. It can efficiently remove the noise, but also can keep the detail feature of the model.Aiming at the problem of points cloud simplification, after introducing the principle and classification of the simplification and the common used algorithm. a new simplification algorithm based on curvature information is proposed. This algorithm gets the curvature information at any points from points cloud by principle component analysis. Then according to the curvature information, it will simplify the points cloud according to different percentage streamline. It improve the rough model and feature missing problem of the traditional space bounding box algorithm.Aiming at the problem of surface reconstruction of scattered points cloud, we mainly introduce the Poisson surface reconstruction algorithm based on implicit surface and the Power Crust algorithm based on Delaunay triangulation. Then we compare the result model of the two algorithm by experiment and choose the Power Crust algorithm as the surface reconstruction algorithm in this paper. Finally, aiming at the rough model situation, the Loop subdivision algorithm is be used in this paper to smooth the model.Finally, we introduce the function model of the software system developed in this paper and run the software.