A Study On Data Filtering And DEM Extraction Of Airborne LiDAR Point Clouds

The airborne-LiDAR(Light Detection And Ranging) is one of the brand-new technology of aerial/space remote sensing, which is assembled by high-technology of GPS, INS and Laser Ranging Scanner. Owing to its efficiently working and accurately data precision, the LiDAR can quickly acquire 3-D information of geo-spatial, which gives people a new way of observation and a new means of measurement, affords a new approach for mapping and surveying and bring new hope to overcoming problems that aren't settled out by traditional survey mapping ways. In the thesis, three key processes in obtaining DEM are discussed, including pre-process of point clouds, point clouds data filtering and interpolation. Through the tests, we can get some useful and helpful results. The primary works and innovations are included as:1,Firstly reviewed the airborne-LiDAR technique theory, and compared with traditional aerial photogrammetry and INSAR. Briefly analyzed the components and characteristics of point clouds, based on which, the pre-process of point clouds is presented. Accomplished the conversion of standardization LAS format point clouds data, and carried out point clouds noise filtering based on height statistics histogram.2,Briefly reviewed some classical point clouds filtering algorithm, and by analysis and conclusion, the thesis presented the improved data filtering methods of hierarchical moving curved fitting algorithm(HMCFA) and TIN filtering algorithm based on height jump(TINFA). The HMCFA used two threshold values to execute filtering algorithm, including window size and difference of height between observation point and fitting curve. Compared with HMCFA, the TINFA also used two threshold values, including ground features height difference and quantity of neighborhood points, to execute filtering algorithm. Based on the characteristics of discrete data, we introduced the k ? d tree data organization, which is not only convenient for executing filtering algorithm, but also can avoid precision loss through others data organization. The experiment results presented the improved algorithms being adaptive and stable.3,Analyzed and concluded the data interpolation methods, and briefly discussed merit and demerit of these methods. The thesis used Kriging, Shepard, and Local Polynomial data interpolation methods to fill up data empty. Besides, every method is compared. In order to assess the obtaining DEM's quality, we presented a primary data assessment scheme and provided detail assessment indexes. The experiment results showed the assessment indexes are reasonable and feasible.