建立準確的3D物件立體模型在製造工程技術有很大的價值。使用高精準度的自動化掃描設備來獲得3D物件的立體模型需要很高的成本。其他非自動化的重建三維模型方法常需要人工參與，當資料龐大時會非常耗時和費力。本研究企圖以平價的結構光設備和攝影機，發展快速的全自主3D物件立體模型建立技術。
先利用格雷碼結構光法分別取得3D物體幾個方向的三維點陣雲資料，再分別取得各掃描表面積上的SIFT特徵，再針對兩個具部分共同表面的點陣雲使用雙向匹配法和RANSAC演算法找出相同面積內之相似點做兩個點陣雲的接合依據，再利用ICP演算法找出適當的轉換矩陣自動接合該兩個點陣雲。重複執行上述步驟即可創造出一個3D物件的完整三維點陣雲模型。

Generation of a precise 3-D model of an object is rather valuable in engineering manufacturing techniques. Using a high-precision automatic scanning device to render a 3-D model of an object demands a high cost. Other non-automatic methods of construction of 3-D models, require manual manipulation, making it time-wasting and laborious when a large amount of data is involved. This study aims to develop a technique of rapid and fully automatic construction of a 3-D model of an object using a structured light approach with a low cost camera and projector.
First, a method based on Gray code structured light approach is used to obtain the data of the 3-D point clouds of an object in a few directions, and SIFT is utilized to find the invariant features on the surfaces of generated point clouds. Then the point clouds that are located on the overlapping regions of any two of the adjacent surfaces are used to find the transformation matrix using bi-direction matching method and RANSAC algorithm so that the two adjacent surfaces can be merged by ICP. Repeating the steps above can lead to a complete 3-D point cloud model of an object.

[1] 陳上勳，"光學非接觸式三維物體輪廓多面量測之研究"，中原大學機械工程研究所碩士論文，中華民國八十四年六月。
[2] J. Y. Shieh and L. H. Shyu, "Measurement of 3D Profile Using Encodeed Light Pattern," 電腦視覺、圖學暨影像處理研討會論文集, pp. 392-397, 1997.
[3] D. G. Lowe, "Object Recognition from Local Scale-Invariant Features," International Conference on Computer Vision, Corfu, Greece (September 1999), pp. 1150-1157.
[4] D. G. Lowe, "Distinctive Image Features from Scale-Invariant Keypoints," International Journal of Computer Vision, 60, 2 (2004), pp. 91-110.
[5] A. Fischler and R. C. Bolles, Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography. Comm. of the ACM, Vol 24, pp. 381-395, 1981.
[6] P. J. Besl and N. D. McKay, A Method for Registration of 3-D Shapes. IEEE Trans. Pat. Anal. and Mach. Intel. 14(2), pp. 239-256, Feb 1992.
[7] N. G. Durle, J. Thayyoo and V. J. Raso, “An Improved Structured Light Technique for Surface Reconstruction of The Human Trunk,” IEEE Canadian Conference on Electrical and Computer Engineering, 1998, 2: 874-877.

[8] G. Sansoni, S. Corini, S. Lazzari, R. Rodella and F. Docchio, 3-D Imaging Based on Gray Code Light Projection: Characterization of the Measuring Algorithm and Development of A Measuring System for Industrial Applications. App1. Opt. 1997, 36(19): 4463-4472.
[9] 李仁舉、鍾約先，三維測量系中攝像機定標技術，清華(自然科學版), 2002, 42(4): 481-484.
[10] 騫森、朱劍英，”基於改進的SIFT特徵的圖像雙向匹配算法”，機械科學與技術, 2007, 26(9): 1179-1182P.
[11] Meshlab, http://meshlab.sourceforge.net/
[12] PCL, http://www.pointclouds.org/