Research On Data Filtering And Building-Footprint Detection Of Airborne LIDAR

The airborne LIDAR(Light Detection and Ranging) is a brand-new technology of remote sensing, which gives people a new way of observation and a new means of measurement, affords a new approach for terrain mapping and bring a new hope to overcoming problems which aren't settled out with some traditional survey mapping ways. In this thesis, the potential applications in obtaining DEM(Digital Elevation Model),DSM(Digital Surface Model) from data gotten by airborne LIDAR are studied and the capabilities with three-dimensional building modeling from data gotten by airborne LIDAR are studied too and the corresponding key techniques of processing data including data filtering and building modeling are analyzed and studied. A system with nealy all functions for data processing is finished on our own which is named by Lidar-Fire.The primary works and innovations are included as:1. The development of the system of airborne LIDAR and its data proessing are reviewed. A classic software named by TerraScan for processing data from airborne LIDAR is introduced and based on which the our deficiencies in proessing data from airborne LIDAR are pointed out. The basic components of airborne LIDAR are described and its working principle is expatiated. The factors which cause errors in source data are analyzed and the models for correcting errors are given.2. The arithmetics of wavelet filter, hierarchy robust adjustment filter and a progressive morphological filter are experimented with data in Germany and Britain and based on which the arithmetic of POBL (Progressive One- dimensional and Bidirectional Labelling) filter is setted forth which is improved on OBL(One- dimensional and Bidirectional Labelling) fileter. With POBL filter, we apply many criterions in the processing and filter data under the complex environments and resolve problems which are encountered with other filter in existence.3. A new measure named neighbor relations is put forward to pick up footprints of building and as a result points on vegetation surface are effectively separated and eliminated from points on building surface. The r- radius point density is introduced to choose seed point which lie in each building roof and as a result the perform efficiency in data segmentation is advanced.4. A new building modeling with tangent rectangular is put forward with which the demand and dependences on the quality of data are reduced.5. A arithmetic of building modeling with data driven way is experimented and based on which, a hybrid building modeling way is setted forth with which both arithmetics of data driven way and model driven way are utilized to rebuild different structure building.6. A software system with nealy all functions for data processing is finished which is named by LIDAR-Fire. Several blocks of data in Narrode Germany and Birmingham England are processed with LIDAR-Fire and TerraScan respectively and their results are compared to each other. As a result, it is indicated that the LidarProcess System is of stabilization, reliable precision and its velocity of processing data and its correctness aren't worse than advanced softwares in existence.