Research On Filtering And Classification Of Airborne LiDAR Data

Airborne LiDAR (Light Detection And Ranging) system provides a novel technical methodfor the acquirement of3D data with high temporal and spatial resolution. It can obtain3Dinformation without the constraint of sunlight or weather conditions and has many merits suchas all-weather, speedy and accurate. Therefore, it plays a more and more important role in theapplications such as3D city modeling, terrain mapping, environment surveillance and resourceinvestigation. In addition, it indicates the new direction of earth-observation technologydevelopment to some extent. This thesis focuses on the theme of airborne LiDAR datapost-processing, and puts keystone on the study of airborne LiDAR data filtering andnon-ground point cloud classification. The major works and innovations are listed as follows:1．The research status and trend of airborne LiDAR system and the technology ofairborne LiDAR point cloud filtering and classification are summarized, thecomponent and format of point cloud are reached, the characters and data organizationmode of point cloud are analyzed, the method of the spectrum fusion between pointcloud data and image data is expatiated, the overall strategy introduing spectruminformation into the point cloud data filtering and classification is concluded.2．Combining with the existing airborne LiDAR point cloud filtering methods, the keypoints in the designing of the airborne LiDAR point cloud filtering algorithms areanalyzed. Then a airborne LiDAR data filtering method based on hierarchicalpseudo-grid concept and surface constraint is put forward. The experiment and erroranalysis validate the availability of the method. Besides, the point cloud is pre-filteredbased on echo information and spectrum information before filtering, which improvesthe performance of filtering.3．A non-ground point cloud classification strategy, which integrates height textureinformation, spectrum information, filtering information and echo information, ispresented on the base of filtering. And then, a classification workflow based onpseudo-grid is designed. Experiment results proved it can manage to classifynon-ground point cloud as buildings point, vegetation point or other non-ground pointsuccessfully.