Research On Simplified Technology Of 3D Scattered Point Cloud Based On Feature Perception

With the development of computer science and virtual reality technology,digital museum technology has gradually become an important technical means for the cultural heritage and natural heritage.The digital museum has the functions of resource sharing and network transmission.If the original scanning point cloud is used for direct modeling,it will consume a lot of time and space resources,increase the burden of network transmission,and reduce the user experience of the digital museum.Therefore,how to simplify point cloud data accurately and effectively has become a key technology for virtual reality display.Based on the above problems,the original point cloud features are preserved in both global and local levels,and the simplified algorithm of 3D scattered point cloud based on feature perception is studied in this paper.The research work of this paper is as follows:(1)Aiming at the problem that the scattered point cloud simplification process easily loses the geometric features of the contour edge,a contour edge feature extraction method based on normal deviation is proposed.Firstly,the point cloud normal vector is obtained by constructing the point cloud spatial topological relationship and local plane fitting.Then,the contour edge point detection method based on coordinate value comparison is used to extract the candidate contour edge points.Finally,a region growth method based on normal deviation is proposed to distinguish the contour edge points and the points near the contour edges effectively.The experimental results show that the proposed method is simple and robust,and can effectively extract the contour feature points of the point cloud model,which has high computational efficiency.(2)Aiming at the problem that the contour shape is rough,the local geometric features are easily lost,and the holes are generated after point cloud simplification,a point cloud simplification method based on expectation maximization clustering is proposed.Firstly,according to the local distribution of points,the point cloud is clustered by the expectation maximization method.Then,the high curvature point is determined by the local absolute curvature in the cluster cluster,and the feature point division function with strong description ability is constructed.Finally,point cloud is simplified based on feature perception Hausdorff distance.Experiments show that the method can be as dense as possible in the place with large curvature,and the position with small curvature is as sparse as possible,while retaining the sharp features and displaying the overall contour as much as possible,effectively avoiding problems such as easy to generate holes during the simplification process and reducing the simplifying error.(3)A three-dimensional scattered point cloud simplification system is designed and implemented in this paper.The main functions include digital management of point cloud model,normal vector estimation,contour edge feature extraction and scattered point cloud simplification.The experimental results show that the system can simplify point cloud data quickly and effectively,and ensure better reconstruction effect.This paper is supported by the National Science Key Project(61731015),“Research on Digital Geometric Virtual Restoration of Damaged Ceramics Cultural Relics” and the National Science and Technology Youth Project(61802311),“Research on Automatic Mosaic Method of Defective Cultural Relics Fragments Based on Deep Feature Model”.