Research On Airborne LiDAR Data Filtering For Steep Area With Dense Vegetation

Airborne Li DAR is a new active remote-sensing measurement technique. The technology can collect 3D spatial geometry information and remote sensing images of a large area effectively. It is widely applied in forestry, power industries, highway reconnaissance and design, coastal engineering, disaster and environmental monitoring, and so on. With the improvement of Global Position System(GPS) and Inertial Navigation System(INS) position accuracy,generating digital terrain model(DTM) from Airborne Li DAR data had become more and more popular. Penetrability of laser beams makes it become an effective way to get the DTM in forest land.However, the measurement of Li DAR is a kind of blind measurement. It collects all the returned pulses from vegetation, birds in flight, man-made objects, etc. Filtering and classification are necessary steps in data processing procedures. The filtering is the process that divided Li DAR data into terrain and off-terrain datasets, which is critical in providing a dataset to develop accurate surface models.The existing filtering algorithms are only suitable for Li DAR point cloud data with flat terrain,sparse vegetation. For area with complex landform covered by dense vegetation, laser data from ground is extraodinary less, hence, Li DAR data filtering has been one of difficult data processing problems that needs further research.This thesis aims at solving the Li DAR data filtering problem in dense forest area based on the multi-pulse information. The main work in this dissertation is as follows:①Considering the great amount of Airborne Li DAR point cloud data, based on the multi-pulse information characteristics, part of the vegetation in multi-pulse is excluded in data processing which can reduce the data volume to some extend;②Selecting the filtering seed points from the single-pulse and last-pulse in two level to increase filtering reliability, taking undetermined point as the center, filtering by searching ground points within a certain range which avoid the iterative operations and cumulative errors;③Based on summarizing the existing filtering methods, putting forward a new filtering way which combines with the minimum distance and the weighted height, reducing the computational complexity effectively, improving the filtering efficiency;④Comparing the results of the experiment with both the original point cloud by visual check and reference data by the quantitative analysis to evaluate algorithm accuracy.