Point Cloud Reduction Algorithm Based On Poisson Distribution And K-means Clustering

With the development of times reconstruction technology based on 3D point cloud has become the focus of research in the hot field of game and film making,computer graphics image processing and reverse engineering,and is widely used in the fields of cultural relic repair and town planning.3D laser scanners can quickly acquire point cloud data which is very large,but the data bring great challenges to the calculation and processing of civil computers,resulting in waste of storage space and calculation.Therefore,it is very important to rationally streamline the point cloud model.The point cloud data reduction algorithm based on surface fitting can preserve detail features of the point cloud raw data,and this method is simple and accurate but the data needs to be segmented firstly,and holes are easily generated in flat parts.The point cloud data reduction algorithm based on K-means clustering algorithm can decompose point cloud data into independent clusters,and can quickly simplify data,but it is not sensitive to the features of the point cloud model.This method is easy to lose feature information,and holes are easily generated in flat parts.The point cloud feature point detection algorithm based on Poisson distribution can accurately detect the feature points in point cloud and avoid the hole phenomenon in the flat parts.However,the algorithm cannot simplify the data of the nonfeature points.In conclusion,this paper proposes a point cloud data reduction algorithm based on weighted least squares curvature calculation and Poisson distribution K-means clustering algorithm.Firstly,the combination of data fitting algorithm based on weighted least squares and point cloud data outlier detection algorithm improves accuracy of curvature calculatio n,and uses this algorithm to calculate the curvature of point cloud.Then,a point cloud data reduction algorithm based on Poisson distribution and K-means clustering algorithm is proposed.According to the results of weighted least squares curvature calculation,the algorithm combines point cloud data reduction algorithm based on K-means clustering and feature point detection algorithm based on Poisson distribution.This algorithm not only can solve the problem of hole in point cloud model,but also ensures the accuracy of the streamlined point cloud data.The main research work of this thesis includes in-depth analysis of existing algorithms and related technologies,and expounds the research status and research hotspots of point cloud data reduction at home and abroad.By learning the related technology of 3D point cloud data reduction,combined with point cloud curvature calculation algorithm and point cloud data segmentation calculation and feature detection algorithm,the retention characteristics of point cloud data are simplified.The innovations in this paper are summarized as follows:1.For the inaccuracy of 3D point cloud data curvature calculation,a curvature calculatio n algorithm based on weighted least squares method is proposed.This algorithm uses outlier rate as the weight of point,and fits clusters of the point cloud data.This method can reduce the influence of outliers on surface fitting,and it can calculate curvature according to the surface equation.Since the outlier rate of the point cloud data is used as the weight,the degree of fitting of the surface is improved,so that the curvature calculation accuracy is higher.2.For the low precision of the point cloud model caused by the 3D point cloud data reduction algorithm,a point cloud data reduction algorithm based on Poisson distribution K-means clustering is proposed.The algorithm uses a curvature calculation algorithm based on weighted least squares method,and according to the threshold of curvature,it is re-clustered in the region with large curvature and decomposed into two smaler clusters.It is decomposed until the curvature of the curvature of cluster is less than the threshold.Then,using the feature point detection algorithm based on Poisson distribution to retain more data points.In summar y,the algorithm can retain more data points,improve the accuracy of the model,and avoid the phenomenon of holes.