A Research On Filtering And Classification Of Airbome LiDAR Point Clouds

Airborne LiDAR system is an emerging surveying and mapping technology. Thispaper systematicly summarizes and anlysizes the basic ranging principle of airborneLiDAR system. Noise point clouds are filtered based on height histogram fromLiDAR point clouds and relationships between point clouds. Three filteringalgorithms-Parameter Improved Morphological method, Binary Quadric SurfaceCombining with Morphological method and Window Iterative Kriging method-areproposed to classify terrain points. Test results show these three methods presenthigher precisions. Building point clouds and its facets are separately classified andidentified by CUMT_PC method and CUMT_AT method. Experimental resultsreveal that classified precisions of building point clouds using both methods canreach precisions above average levels; detection percentage and quality factor ofthese two methods are higher than92%which meet the requirements of automaticidentification. According to height and intensity of LiDAR point clouds, initial roadpoint clouds of urban region are distinguished and a method of extracting urbanvector main road network is developed. Trial results display that completenesspercentage and correctness percentage of extracted roads are good.