Research On Lidar Data Filtering Method Using Parallel Platform

Airborne LiDAR technology has gradually developed into an important means of getting high effectiveness and high resolution spatial data it is also widely applied in mapping, resource investigation and environmental protection. Compared to traditional methods of measurement, LiDAR data system has more accurate elevation resolution, shorter cycles and higher degree of automation. The most direct application of LiDAR data lies in generating digital elevation models from point cloud data, with data filtering being one of the most important links. The huge amount of point cloud data increases the difficulty to filter calculation. Many filtering algorithm with qualified theoretical basis face the problem of running too long when dealing with a wide range of LiDAR data. In this case, parallel Computing provides an efficient platform for data-intensive tasks.This dissertation aimed to study the programming of data filtering algorithm based on message passing interface parallel computing platform. The main contents include:1. Based on LiDAR data, the dissertation discusses the working principle and technical characteristics of airborne LiDAR system, conducts studies on LiDAR data structure, and compares the similarities and differences of the LiDAR, aerial Photogrammetry as well as synthetic aperture radar data;2. The dissertation studies on the principle of LiDAR data filtering and analyses several common filtering algorithm for point cloud data, which focuses on mobile filtering algorithm and offers improvement proposals--that is eliminating low gross errors in the data before determining the initial fitting surface;3. The dissertation analyses the controlling structures and communication model of parallel computing platform. It specifies a detailed study on the normal processes design of parallel calculation algorithm and focuses on message passing model, including the principles and methods of message passing interface programming;4. The empirical part of this dissertation adopts the example of mobile LiDAR data filtering surface fitting method and utilizes message passing interface MPI to make parallel programming. It divides the massive LiDAR data into blocks and adopts master-slave mode to design the dual-process and the process of four parallel programs. It also makes comparative analysis with serial program of the algorithm;The result of this dissertation supports the author’s hypothesis that parallel computing is more superior to other platform in processing algorithms for airborne LIDAR data. Although the newly improved mobile plane filtering algorithm has some erroneous problems, they do not greatly affect the algorithm. In contrast, the new algorithm is much more efficient in LiDAR filtering.