Spectral graph based approach for analysis of 3D LIDAR point clouds

Thesis Type: Postgraduate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Electrical and Electronics Engineering, Turkey

Approval Date: 2017




Airborne Laser Scanning is a well-known remote sensing technology, which provides quite dense and highly accurate, yet unorganized, point cloud descriptions of the earth surface. However, processing of such a 3D point cloud is quite challenging due to its irregular structure and 3D geometry. In this thesis,two novel approaches for the analysis of unorganized 3D point cloud data are proposed, specifically the ones that are generated by the airborne mounted LIDAR sensor. These methods rely on the spectral graph based and graph signal processing techniques which gain attention in the recent years. The state-of-the-art techniques addressing the problems of LIDAR point clouds are first examined. Next, the theory presented by the spectral graph based methods is reviewed to analyze their solutions. Since irregular discrete data lying on a high dimensional geometry, such as LIDAR point clouds, can be conveniently represented by weighted graphs, spectral graph methods based on such weighted graphs enable spectral analysis of the data representation, as in classical Fourier analysis for signal processing. In the light of the revisited spectral graph literature, one can examine techniques for clustering as well as edge detection problems by using graph representation of the unorganized 3D point clouds. The graph based representation introduces the opportunity of analysis of the signal over its original input space; therefore, it provides qualified comprehension of the data. Based on simulations, it is shown that the graph spectral solutions can acquire remarkable advance in the analysis of unorganized 3D point clouds and the experimental results indicate the potentials of this new approach.