Research On Point Cloud Compression Technology Based On Spatial Index

Point cloud data is an important kind of three-dimensional data.With the continuous development and maturity of the technology of rapid point cloud acquisition,the data volume of point cloud is also increasing rapidly,and effective data compression methods are urgently needed.Common point cloud compression methods include lossless compression and point cloud reduction.This paper studies the lossless and lossy compression of point cloud based on spatial index,k-nearest neighbor search and other methods.Lossless compression is a general compression method.Information entropy theory indicates that lossless coding compression has a certain compression limit.Statistical coding is an important method of lossless compression,among which arithmetic coding can achieve the compression effect close to information entropy.As scattered point cloud data has a high degree of spatial correlation and disorder,it is possible to make the data more orderly through a reasonable organization method without changing the coordinate accuracy.Common point cloud organization methods include k-d tree,octree,R tree,regular grid index,space filling curve and other spatial index structures and extended index structures based on these methods.Point cloud simplification methods include random simplification,uniform simplification and feature preserving simplification,etc.Before or during simplification,outliers and noise points are generally removed.In the simplification method,the feature preserving simplification method has a large amount of computation,but because it has the advantage of preserving the details of the original point cloud,it has attracted extensive attention.This kind of method generally needs to obtain the neighborhood information of point cloud.k-d tree and octree are often used in nearest neighbor search.Approximate nearest neighbor search can also be used when accuracy is not strictly required.The curvature of point cloud at a certain point is an important index of point cloud simplification.In the practical application of point cloud data,the method of surface reconstruction is often used.The surface reconstruction is generally based on the calculation of the isosurface,and the commonly used methods include moving cube method,Poisson method and greedy projection triangulation method.The approximation between surfaces can be measured by the Hausdorff distance.The specific research content of this paper is as follows:(1)Make full use of the characteristics of point cloud data,put forward a point cloud compression method without loss of precision,this method uses linear octree and Morton code to organize and code point cloud,and then adopts and improves several statistical coding methods to achieve a relatively ideal compression rate.This paper also studies the influencing factors of the compression rate of this method,and comes to the conclusion that the compression effect of this method is more ideal when the point cloud density is high and the accuracy is low.(2)the performance and influencing factors of neighborhood search for k-d tree and octree are studied,and on this basis,a point cloud simplification method based on curvature and point distance is proposed.The point cloud obtained by this method is evenly distributed and has a strong ability to maintain details.(3)several common surface reconstruction methods were introduced and implemented,and the greedy projection triangulation method with the best performance was selected to reconstruct the original point cloud and the reduced point cloud of Happy Buddha,and the model before and after the reduction was compared using the Hausdorff distance method.Experiments further verify the effect of point cloud simplification.