Research On The Filtering Algorithm Of 3D Laser Scanning Point Cloud Data

The 3D laser scanning technology has been widely used in 3D surveying and mapping.The acquisition of point cloud data will be deeply influenced by instruments,the environment and the surveyors,resulting in many noise points.The data including these noise points will have a huge impact on the reconstruction of point cloud models.Therefore,in order to obtain the high-precise point cloud model,the point cloud data must be filtered to remove these noise points in the point cloud.The purpose of point cloud filtering is to try to remove the noise points that are deviated from the surface,to keep the point cloud of the object surface,and to keep the obvious features and the characteristic information of the geometric edges of the scanning model surface.The paper studies and analyzes the filtering process of scattered point cloud data preprocessing in depth.The key points are as follows:1?Because of the unobvious topological relation of scattered point cloud,the paper proposes the topological relation of scattered point cloud by using octree method,and the K-neighborhood search is introduced based on the introduction of common point cloud data.Aiming at the problems of large amount of search data,incomplete search,and low search efficiency for the direct search method,a hierarchical search method is proposed to search from the inside to the outside to improve search efficiency.At last,the average value of the K adjacent points acquired by the neighborhood mean value method is obtained and the data is compared with the given threshold.The experiment shows that the neighborhood averaging method has achieved good results.2?Based on the introduction of fuzzy mathematics and Fuzzy C-Means clustering,according to the features that FCM clustering is not sensitive to point cloud denoising and small noise points,the covariance factor is introduced to increase the weight of fine noise.Aiming at the issue that clustering center has huge influence on filtering effect,the paper proposes Mahalanobis distance instead of Euclidean distance in FCM clustering,and then derives the formula of fuzzy C mean clustering algorithm based on Mahalanobis distance.The experiment shows that the clustering center of FCM-M algorithm is closer to the surface of the model,which means Mahalanobis distance is better than the classical FCM algorithm and has good smoothing effect on the point cloud model.3?On the basis of qualitative analysis of the experimental filtering effect,as for the issue that qualitative analysis is not intuitive enough,the effect of filtering algorithm is analyzed quantitatively by method noise.The experiment shows that the FCM-M algorithm has good denoising effect and smoothing effect on the point cloud data and the feature information remaining of the model.