Study On Trees And Buildings Extraction From Urban Area Based On Airborne LiDAR Point Cloud Data

Urban buildings and trees extraction based on airborne Li DAR data has been the research hot pot in LiDAR technology application, so the deep study on it has an extremely important practical significance and application value. But due to the characteristics of airborne LiDAR point cloud data and the complexity of the urban environment, the research about the tree extraction from LiDAR cloud data relatively lacks. In addition, the algorithm of building extraction using airborne LiDAR data is already very mature, but there is little study on the object-based building extraction based on airborne Li DAR data alone, mainly focus on the airborne Li DAR data used as an auxiliary data for remote sensing image. Therefore, based on the analysis of characteristics of the airborne LiDAR system and data alone, this paper focuses on the filtering methods of Li DAR cloud data, then further discusses the method of trees extraction based on the difference of twice return pulse heights or intensity, under the premise of these results, an strategy of object-based building extraction and a method of trees extraction are bringed out. The details are as follows:(1)Iterative TIN filtering algorithm is analyzed emphatically, aiming at the higher computational complexity of the algorithm, an improved iterative TIN filtering algorithm for LiDAR data is made by region growing growing of seed points and filtering result, in order to leverage the respective strengths of the first and last return pulse data, the first return pulse data is used to select seed points and the last return pulse data is used for other filter processings, at the end, an experiment has been done.(2)Trees extraction from airborne Li DAR data is deeply studied. Firstly, in order to deal with the problem that the building boundaries misclassified as trees easily in the method of trees extraction based on difference of twice return pulse heights, a method is proposed which uses the slope to locate the edges and uses the intensity to distinguish the buildings at the edges, the experimental results show that the approach is valid and robust. A method to extract trees using the difference of twice return pulse intensity is given, validation test shows that the method trees extraction has got good effect.(3)An strategy of object-based building extraction is bringed out. Firstly, the DTM is generated by interpolating the Li DAR ground point, and then the nDSM which is calculated by stracting the DEM from the DSM is used to extract buildings, so it can ensure the process is almost not affected by topography; The nDSM image is segmented into trees area and non-trees area based on the result of the trees extraction, and then the multiresolution segmentation algorithm is used to segment the non-trees area in the nDSM with edge information from canny detector, further, the segmentation result is optimized through merging the adjacent objects with smaller height difference, finally the intensive buildings are extracted based according to the spectral, geometry character and the spatial relations of objects. We compares with the two results from two aspects which are visual impact and precision assessment, it is depicted that the extraction of buildings from the method based on object-based is far superior to the method based on pixel-based.