Aiming at the problem of multi-scale noise encountered in the actual acquisition of 3D point data model,a method based on noise classification combined filter smooth de-noising is proposed. Firstly,according to the characteristics of the position,density and uselessness of the noise in the point model, it is divided into three categories:pseudo noise,large scale noise and small scale noise. By using direct filtering,radius filtering and sta-tistical filtering,and moving least squares re-sampling to remove. And the combination of the VoxelGrid filtering and moving least squares re-sampling is taken into account,and the optimization of the re-sampling point model is realized. The experimental results show that this method can solve the problem of multiple scale noise in the point data model collected by the 3D laser scanner. From the loopholes in the visual effect of the model,the feature re-tention and the smoothing effect and the actual deformation,this method is superior to the direct use of MLS re-sam-pling and Laplace smooth de-noising.%针对在实际采集三维点数据模型中遇到多种尺度噪声的问题,提出了一种基于噪声分类组合滤波平滑去噪的方法.首先根据点模型中噪声的位置、密度和无用性等特点,将其分为伪噪声、大尺度噪声和小尺度噪声三类;分别使用直通滤波、半径滤波和统计滤波,以及移动最小二乘重采样(MLS)去除;同时考虑到点云冗余,将体素化网格滤波与移动最小二乘重采样相结合,实现重采样点模型的优化.实验结果表明:能够很好的解决地面三维激光扫描仪采集的点数据模型中存在多种尺度噪声的问题.从模型视觉效果上的漏洞修复、特征保持和光顺效果,以及实际变形度上看,方法优于直接使用MLS重采样和Laplace平滑去噪.
展开▼
