Research Of Segmentation Algorithm For Scattered Point Cloud Based On Local Convexity

With the continuous development of three-dimensional laser imaging technology in recent years,three-dimensional imaging system is widely used in reverse engineering,digital city,intelligent cars and so on,followed by a higher and hiter requirements for the accuracy and speed of point cloud data processing.In order to meet the needs of high-precision data processing results in practical application,point cloud data processing has become research focus for domestic and foreign institutions.The data collected by the three-dimensional laser imaging system is a large number of scattered point cloud data.First,the point cloud data should be divided,and the objects with different attributes should be separated to provide basic information for subsequent operations.Then,the following procedure can be classification,recognition,reconstruction and other processing.As cloud point segmentation is a main part of point cloud data processing,it is very important whether the results of segmentation and segmentation can meet the practical application requirements.Based on the analysis and summarization of the problems in the existing algorithms for dealing with the actual point cloud data,this paper intends to use the improved local convexity algorithm to segment the outdoor complex point cloud data,and verify the experiment.The main contents of this paper are as follows:1 ? This paper introduces the point cloud data processing.Furthermore,it hignlights the importance of segmentation in autonomous driving system and robotics system.And the results of segmentation is the information support of the two systems.2?Based on the introduction of the min-cut based segmentation and the region growing segmentation,the algorithms are experimented by scattered point cloud data in complex environment.The results of experiments show that the two algorithms can segment regular object effectively.Nevertheless,the experimental results of the irregular object's segmentation are not perfect.3?Aiming at the problem of oversegmentation and undersegmentation in the segmentation experiment,an improved local convexity segmentation is proposed.The algorithm replace the triangulation in the original algorithm by constructing a local connected point set and segment point cloud data by connected components labeling.The algorithm is experimented by scattered point cloud data in complex environment.The results of the experiments show that the improved local convexity segmentation of connected point sets is better for segmentation of actual road information.The results also prove that this method is suitable for segmentation of scattered point cloud data in complex environment and the more the neighbor points,the better the segmentation results.