Study On The Technology Of Feature Extraction And Filtering For Airborne LiDAR Points Cloud

Three-dimensional laser scanner technology is a science and technology innovation in the field of Surveying and mapping and other fields,such as machinery,architecture and so on.The method of spatial information acquisition based on airborne LiDAR technology has shown great advantages and potential in many industries.Aiming at the problem of point cloud data processing for airborne LiDAR In this paper,we respectively from the points cloud error elimination,point cloud filtering,point cloud feature extraction,point cloud classification and application modeling and other aspects of the research,through the experimental test to complete and comparison the algorithm.And especially focuses on the problem of gross error elimination,filtering and feature extraction of airborne LiDAR point cloud.The main work and achievements are as follows:(1)Due to the characteristics of mass and scatter of airborne Li DAR point cloud,a case study of searching k neighbor points,testing the importance of point cloud data organization for point cloud data processing;On the premise of constructing k-d tree,we test the relationship between different k neighbor points and time cost in different samples.(2)Through the actual data experiment,the detection method based on the frequency distribution,the grid information and the point cloud rough approximation based on the neighboring information is completed by the ratio of the ground points in the culling points,to obtain the empirical values of each method in practical application.(3)A method for airborne LiDAR point cloud filtering based on multi view projection from the view of automation and segmentation was puts forward,not only introduce the principles,but also completed the test by an example.(4)The feature extraction software of point cloud is designed and developed,taking the method of 3D rendering shader from different types of features including elevation,curvature and spatial characteristics are extracted from the simulated data,analysis the applicability of the various features and related parameters set reference value.(5)Taking the measured point cloud data of natural landform and city as an example,on the premise of completion the relevant feature extraction,to realize the extraction of the main feature points in the contour line and the ditch of the natural landform point cloud.Especially focus on urban experimentation area,to analysis the effect and limitation of line features and surface features extracted from the experimental area.Then,puts forward using the normal vector Angle change degree,relative elevation and other characteristics,adopt the method of constructing rule set classification step by step to realize the classification of point cloud,and experiment test;Finally,from the perspective of classification efficiency and automation lead into the SVM,and test their classification effect.