Research On Fast 3D Model Rendering Technology Based On Point Cloud

In recent years,with the increasing popularity of engineering applications in the fields of computer graphics,virtual reality,machine vision and reverse engineering,point cloud-based model reconstruction has become more and more widely used,and related technologies have also been rapidly developed: Point cloud equipment acquisition accuracy is improved,point cloud acquisition method is more convenient,and point cloud data also shows a quantitative growth trend.In this context,how to quickly construct a point cloud topology,how to improve the point cloud search speed,how to quickly complete point cloud reduction on the premise of maintaining feature details becomes the core research problem of model reconstruction.This paper focuses on the rapid modeling of spatial scattered point clouds,and deeply studies key technologies such as point cloud topology construction,point cloud denoising,point cloud reduction and point cloud gridding,and implements a point cloud data rendering system based on this design.In the research process,three kinds of point cloud topology construction algorithms,such as spatial cell,octree and KD tree,are introduced.Based on this,K-nearest neighbor search is implemented.Because this method is a serial processing method,it is not suitable for large-scale points.Cloud search problem,so this paper introduces CUDA acceleration in K-neighbor search,and realizes parallel search;in terms of point cloud denoising,this paper uses DBSCAN method to quickly eliminate the noise and external points in point cloud;point cloud reduction,first introduce,the current mainstream point cloud feature point extraction method and point cloud reduction method are used to analyze the influence of different streamlined methods on the modeling effect.On this basis,the advantages of random point cloud simplification method are extracted,and a PCA feature point extraction is given.The point cloud hybrid simplification method realizes the rapid simplification of the point cloud without losing the special feature details.In the aspect of mesh generation,this paper adopts a grid model construction method based on partition DT and GPU parallel acceleration.The point cloud dynamic partitioning strategy realizes the automatic balancing of the division and improves the efficiency of model construction.In order to verify the feasibility and effectiveness of the above method,this paper uses JavaWeb development technology to design and implement a point cloud data rendering system.By loading point cloud data online,it can quickly complete the construction of point cloud data topology and point cloud data denoising.The steps of point cloud reduction and model construction finally realize the model reconstruction of scattered point cloud data.The experimental and test results show that the point cloud processing related methods and the completed point cloud rendering system given in this paper have achieved the expected results.