Research On 3D Point Cloud Filtering And Simplification Algorithms

In recent years,with the development of 3D scanning measurement technology,3D point cloud data has been widely used in fields such as industrial measurement,autonomous driving,medical industry,and urban planning due to its convenient acquisition and strong expression ability.Under such a huge demand,higher requirements have been put forward for the quality of 3D point cloud data.In practical applications,point cloud data obtained through 3D scanning measurement instruments is often massive and mixed with a large amount of noise.These data must be processed before being applied to various fields.Therefore,it is necessary to study the problems in 3D point cloud data processing.This article mainly focuses on the filtering and simplification of 3D point cloud data,and the specific research content is as follows:Firstly,in terms of filtering 3D point cloud data,in order to address the issue that traditional point cloud filtering algorithms cannot completely remove noise of different scales from point clouds,this paper proposes a combined filtering method that improves local anomaly factors.This method combines the average distance threshold with local anomaly factors,first removing large-scale dense noise points through the average distance threshold,and then using local anomaly factors for further outlier detection,Improved the algorithm’s ability to detect and remove large-scale noise points.Finally,a point cloud guided filtering algorithm was used to smooth the denoised point cloud model,removing small-scale noise points from the point cloud model,and achieving rapid denoising and smoothing of noise at different scales.Secondly,in terms of 3D point cloud data reduction,this paper proposes a point cloud reduction algorithm based on normal angle region division to improve the traditional principal component analysis method for calculating normal vectors by adding reciprocal distance square weights to improve the calculation accuracy of point cloud normal vectors,as traditional point cloud reduction algorithms cannot fully preserve model features and are prone to point cloud holes caused by over simplification in non feature areas,Afterwards,the average normal angle of the point cloud is calculated,and the point cloud is divided into feature regions and non feature regions using the average normal angle threshold.By setting different minimum bounding box edge length thresholds,uniform and effective simplification of different regions of the point cloud is achieved.Finally,the point cloud of the workpiece mold and the point cloud of the ceramic plate obtained by the surface structured light 3D scanning measuring instrument are taken as the experimental objects,and the algorithm in this paper is compared with the commonly used point cloud filtering and reduction algorithm to verify the effectiveness of the algorithm.The experimental results show that the proposed filtering algorithm can effectively remove noise of different scales,and the denoising efficiency is improved by about 23% compared to statistical filtering algorithms.The proposed reduction algorithm can still effectively preserve the contour feature information of the point cloud model at a high reduction rate of 80%,and the maximum and average errors of the two models before and after reduction are smaller than the other three algorithms.