3D Reconstruction Based On Data From Laser Range Finder Of Mobile Robot

With the recent development of 3D data collection technologies and hardware facilities such as laser, with the recent development of robot control technology and the computer technology, 3D reconstruction is widely used. Tnangulation of scattered data is an important step and the base of later step of reconstructing curved surface in 3D reconstruction. After backward glance and summary of theory and method of 3D reconstruction, this paper makes experiments in order to get 3D reconstruction from laser data of mobile robot.First of all, in the aspect of 2D triangulation, this paper summarizes the foundational theories of 2D triangulation in detail, and proposes the viewpoint of combining the method of divide-conquer and add-point algorithms, which has the advantages of high efficiency and small occupied memory. This dissertation obtains triangulation with data of roof. And through simulation study with different dividing threshold value, the paper acquires relationship of dividing threshold value and efficiency of algorithm and examines the effectiveness of the combined algorithm.Secondly, as to constrained triangulation, this thesis modifies existent algorithm, which inserts constraint line segments after establishing initial triangular mesh. The new algorithm decreases the searching burden by blocking and indexing point set, edge and triangle, and reduces compute complexity for it doesn't need to combine every section. The new algorithm also manages the points of influence domain so effectively that it increases efficiency. This dissertation obtains triangulation with data of corridor, and analysis the main reason of this algorithm.Finally, with regard to triangulation from 3D scattered points, this essay establishes contours, uses and compares min-diagonal method and control-points-match method, on considering the characteristics of experimental data. This paper also analyzes the reason that the latter method is better than the former when the adjacent contours are not similar during experiment. The dissertation uses interpolation function instead of line segment as well, when reduce the amount of scattered points.