臺灣博碩士論文加值系統

隨著三維運算技術及其應用迅速的發展，對物件形體之數位化工作的需求與日俱增。相較於小型物件，大型物件的三維重建工作在資料量、精確性上都極具挑戰性。雷射測距掃瞄為一種快速而精確的三圍重建方法，其快速掃瞄大片面積、不易受環境影響的高精確性，使其成為大型物件三維重建工作的首選方法。本文使用一具高精度的Velodyne雷射掃瞄儀，配合ICP剛性對齊演算法，建立了一套快速有效的大型物件重建系統。ICP(Iterative Closest Point)演算法被廣泛的應用於三維重建和機器人定位工作上(SLAM)，是剛性對齊的主導性演算法。本文對原始ICP進行實作改良，使用最差忽略(Worst rejection) 、特徵擷取(Feature extraction)以及跳圈等方式，大幅降低ICP演算法的累積誤差，使其能夠有效的對雷射掃瞄取得的大量資料進行對齊，最終建立精準的大型物件模型。

With the rapid development of 3D computing technology, the acquisition of range data has become a necessary activity for numerous applications. Large scene reconstruction, which aims to gather depth information of the environment by the use of some range sensors, is a challenging problem. It is considered challenging because the scale of scanned data is relatively large than any regular sized object.
Laser rangefinders are perhaps the most frequently used sensors in the applications of scene reconstruction. In our work a 3D reconstruction system that equips with a Velodyne LIDAR is built. In order to improve the range and the density of reconstruction the system implements a modified Iterative Closest Point (ICP) algorithm. The combination of various strategies such as worst rejection and feature extraction is proposed to achieve significant improvement. The experiment shows that the heuristic registering algorithm is capable to align multiple LIDAR scans that consist of large number of 3D coordinates in a faster and more accurate manner compared to conventional ICP algorithm.

中文摘要
Abstract
謝 誌
目 錄
圖目錄
第一章 緒論
1.1 前言
1.2 研究目的
1.3 研究方法
1.4 論文架構
第二章 三維重建技術
2.1 接觸式掃瞄
2.2 非接觸式被動掃瞄
2.2.1 立體光學法(Photometric stereo)
2.2.2 輪廓成型法(Shape from contours)
2.2.3 雙眼視覺法(Binocular stereo vision)
2.3 非接觸式主動掃瞄
2.3.1 結構光法(Structured light)
2.3.2 調變光法(Modulated light)
2.3.3 雷射測距法(Laser rangefinder)
第三章 ICP(ITERATIVE CLOSEST POINT)演算法
3.1 剛性對齊(RIGID ALIGNMENT)
3.2 原始ICP演算法
3.3 ICP對齊圖解
第四章 系統架構
4.1 系統流程
4.1.1 Velodyne HDL-64E 雷射掃瞄儀
4.1.2 實驗環境
4.1.3 轉換掃瞄儀網路封包
4.1.4 近遠點刪除與環狀干擾
4.1.5 ICP對齊與前導位移矩陣
4.1.6 三維重建與機器人定位
4.2 原始ICP演算法與累積誤差
4.2.1 採樣誤差
4.2.2 視野遮蔽
4.2.3 累積誤差
第五章 改良ICP
5.1 最差忽略法
5.1.1 定義忽略
5.1.2 忽略比例
5.1.3 最差忽略法流程
5.1.4 最差忽略範例
5.1.5 最差忽略ICP實驗結果1
5.1.6 最差忽略ICP實驗結果2
5.2 特徵擷取法
5.2.1 特徵攝取法流程
5.2.2 八叉樹
5.2.3 取特徵法實驗結果1
5.2.4 取特徵法ICP實驗結果2
5.3 跳圈法與累計誤差
5.4 實驗結果討論
第六章 結論與未來研究方向
6.1 結論
6.2 未來研究方向
參考文獻

[1]C. Dorai, J. Weng, and A. Jain, “Registration and Integration of Multiple Object Views for 3D Model Constrution,” IEEE Trans. PAMI, Vol. 20, No. 1, pp. 83-89, 1998
[2]C. C. Wang, Simultaneous Localization, Mapping and Moving Object Tracking, PhD thesis, Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, April 2004
[3]H. Seungpyo, K. Heedong, K. Jinwook, “VICP: Velocity updating iterative closest point algorithm,” Proc. ICRA, pp.1893-1898 , 2010
[4]J. Diebel, K. Reutersw?鑼d, S. Thrun, J. Davis, and R. Gupta, “Simultaneous Localization and Mapping with Active Stereo Vision,” Proc. IROS, vol. 4, pp. 3436–3443, 2004
[5]J. Luck, C. Little, and W. Hoff, “Registration of Range Data Using a Hybrid Simulated Annealing and Iterative Closest Point Algorithm,” Proc. ICRA, pp.3739-3744 , 2000
[6]J. P. Godbaz, M. J. Cree, and A. A. Dorrington, “Undue Influence: Mitigating Range-Intensity Coupling in AMCW ‘Flash’ Lidar using Scene Texture,” Proc. IVCNZ, pp. 304-309, 2009
[7]K. Pulli, ”Multiview Registration for Large Data Sets,” Proc. 3DIM, pp.160-168, 1999
[8]M. Ribo, M. Brandner, “State of the art on vision-based structured light systems for 3D measurements,” Proc. ROSE, pp. 2-7, 2005
[9]P. J. Besl, N. D. McKay, “A method for registration of 3D shapes,” IEEE Trans. PAMI, Vol.14, No. 2, pp 239-254, 1992
[10]R. J. Woodham, “Photometric methods for determining surface orientation from multiple images,” shape from shading, MIT Press, pp. 513-531, 1989
[11]R. Klette, K. Schluns, and A. Koschan, Computer Vision：Three-Dimensional Data from Images. Springer, 1998
[12]S. D. Cochran and G. Medioni, “3D surface description from binocular stereo,” IEEE Trans. PAMI, Vol.14(10), pp. 981-994, 1992
[13]S. Rusinkiewicz, M. Levoy, “Efficient Variants of the ICP Algorithm,” Proc. 3DIM, pp. 145-152, 2001
[14]T. Masuda, K. Sakaue, and N. Yokoya, Registration, “Registration and integration of multiple range images for 3-d model construction,” Proc. CVPR, pp. 879–883, 1996