الگوریتم فیلترینگ ابر سه بعدی نقطه ای
Abstract
1-Introduction
2-Related work
3-Proposed Algorithm
4-Experiments and Result Analysis
5-Conclusion
Acknowledgement
References
Abstract
In order to effectively smooth the noise in 3D point cloud without losing the detailed features of the model, a new filtering algorithm based on surface variation factor segmentation is proposed. The method adopts different filtering algorithms on different feature regions of the model. Firstly, the normal vectors of a point cloud model are estimated by using weighted principal component analysis (PCA) method, and the surface variation factor of each point is estimated. Secondly, the point cloud model is divided into flat regions and mutant regions by comparing the surface variation factor of the sampling point with the average surface variant factor of the sampling point k-neighborhood. Finally, the improved median filtering algorithm is applied to flat regions, and improved bilateral filtering algorithm is applied to mutant regions. Experimental results show that it has a preferable smoothing effect and reserves the detail features of point cloud.
Introduction
3D reconstruction technology is widely used in reverse engineering, 3D printing, virtual reality, archaeology, medicine and other fields [1-3]. 3D point cloud modeling is an effective objects modeling method, however, the premise of reconstructing the scanned object is to obtain the real data of the object surface. But because of some human or environmental factors, as well the defects of the scanning device itself, unreasonable noise will inevitably exist in the scanning data, however, these noise data will cause serious problems for subsequent related processing in modeling and measurement [4-6]. Therefore, 3D point cloud filtering is a key step before modeling. The purpose of filtering is to effectively eliminate and smooth the noise in 3D point cloud model, and reserve the original detail features of the object surface.