Research On The Key Techniques Of The Airborne LIDAR Data Processing

As a remote sensing technology, LIDAR (Light Detection and Ranging) is widelyused in many areas, for example, agriculture, forestry, geology, military, astronomy,robotics, transportation and so on. It sends a laser plus to the object, then detects the laserreturn waveform, and gets the3D point cloud of the object at last. There are several typesof LIDAR device, GLAS (Geoscience Laser Altimeter System), land-based LIDAR,mobile LIDAR and airborne LIDAR whose carriers are different. Airborne LIDAR is aremote sensing technology based on the laser ranging technology, GPS (GlobalPositioning System) and IMU (Inertial Measurement Unit). Compared with other remotesensing methods, it is widely used because it is low cost, fast, high resolution and nearlyfull automatic.Here are two key techniques in LIDAR data processing. One is Gaussian waveformdecomposition, another is LIDAR point cloud processing.The aim of Gaussian waveform decomposition is to decompose return waveforminto a series of basic Gaussian components. Each Gaussian can be used to calculate theposition of a specific reflecting point.The aim of point cloud processing is to extract useful information from thelarge-scale point cloud data, for example, digital terrain model, the forest cover rate, theheight of trees. The difficulties are the large-scale point cloud processing, classificationand reconstruction and so on.In this paper I will introduce my job in the following area.(1) The Gaussian decomposition initial value estimate and optimization. Thetraditional Gaussian decomposition initial value estimate algorithm is not accurateenough and not robust to random noise. In this paper I develop a initial value estimate algorithm which is robust to random noise. It is able to get the accurate initial valuewithout filtering the waveform first. Then I use LM Algorithm (Levenberg-MarquardtAlgorithm) and L-BFGS(limited-memory Broyden-Fletcher-Goldfarb-Shanno) algorithmto optimize the Gaussian decomposition result, and get the high-precision point cloud atlast.(2) The real time visualization and interactive edit for large-scale point cloud. Idevelop software for LIDAR point cloud visualization and processing. I also use Octreeto speed up the interactive edit for the point cloud.(3) Develop algorithm for extracting useful information from LIDAR point cloud,especially the classification and DTM (digital terrain model).(4) Develop a series of LIDAR data software, including waveform decompositionsoftware LasView and point cloud processing and visualization software LIDARPC. Inthis paper I will introduce the development of this software.