現今自動化生產線上的零件組裝大都以多台機械及類似機構，重複一樣的路徑去進行零件移動、定位及裝配，此方式相當佔用空間，且須耗費相當多人力、物力及時間進行複雜的前置作業。本研究探討如何在最少前置作業的條件下，能使用單一機械手臂搭配工業相機來進行簡易零件的自動化組裝，其基本想法是先擷取最終組件產品的2D影像，再利用OpenCV函式庫之特徵點擷取、特徵比對及影像旋轉等影像處理功能來辨識各個零件的組裝位置，據以計算零件的安裝順序，以及機械手臂的路徑；接著藉由工業相機擷取散落在桌面上所有零件的影像資訊，再使用Speed Up Robust Feature、K Nearest Neighbor、Homography等演算法進行影像分析，由桌面自動化尋找欲組裝的零件，並求得適當的零件吸取位置及零件的組裝方位；最後使用EPSON機械手臂一一組裝所有零件，使完成的組件實物與原始擷取的組件影像相同。
本研究除了說明影像的特徵點擷取、特徵比對及旋轉校正零件的演算法外，並以積木作為案例，來驗證所開發系統的實用性。

Nowadays, most of the assembly operations of automatic production lines use multiple machines and similar mechanisms to repeatedly move, locate and assemble parts. This type of assembly equipment not only takes up a lot of space, but also needs plenty preparation work for initializing the project of assembling a new product, resulting in high demand of manpower, material resources, time and costs. This study proposes the use of a single robotic arm in together with an industrial camera to complete the assembly task and the procedure requires little preparation work. The basic concept is to capture the 2D image of the final assembly as the target assembly. Based on the assembly image file, some image processing functions of the OpenCV library, such as feature point extraction, feature comparison and image rotation, are used to identify the assembly position of each part. The assembly sequence and the robot arm path are then calculated from the image file of the target assembly. After that, a camera is used to capture 2D images of all parts which are randomly placed on the table. The image files are analyzed using some existing algorithms, such as Speed Up Robust Feature, K Nearest Neighbor, Homography, to find the part to be assembled for each assembly step. The robot arm grasping point and the part’s orientation are also calculated. Eventually, an EPSON robot arm is used to sequentially assemble all parts on the table, and the final assembly is identical to the target assembly image captured previously.
In addition to explaining the algorithms of feature point extraction, feature comparison and image rotation, this study uses a building block to verify the practicability of the developed system.

[1] David, G.L. (2004), “Distinctive Image Features from Scale-invariant Keypoints,” International Journal of Computer Vision, Vol. 60, No. 2, pp. 91-110.
[2] Herbert, B., Tinne, T. and Luc, V.G. (2006), “SURF: Speeded Up Robust Features,” Proceedings of the 9th European Conference on Computer Vision, May 7-13, Vol. 110, No. 3, pp. 404-417.
[3] Jang, H.Y., Moradi, H., Hong, S., Lee, S. and Han, J. (2006), “Spatial Reasoning for Real-time Robotic Manipulation,” Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, October 9-15, Beijing, China.
[4] Lee, S.H., Kim, E.Y. and Park, Y.H. (2006), “3D Object Recognition Using Multiple Features for Robotic Manipulation,” Proceedings of IEEE International Conference on Robotics and Automation, May 15-19, Orlando, Florida, USA.
[5] Hung, S.C. and Li, M.F. (2007), “A SIFT Method Application for a Six DOF Manipulator Grasping an Object Against Complicated Environment,” Proceedings of the 33rd Annual Conference of the IEEE Industrial Electronics Society, November 5-8, Taipei, Taiwan.
[6] Lee, S., Jang, D., Kim, E., Hong, S. and Han, J. (2005), “A Real-Time 3D Workspace Modeling with Stereo Camera,” Proceedings of IEEE International Conference on Intelligent Robots and Systems, August 2-6, Edmonton, Alberta, Canada.
[7] Tomono, M. (2006), “3-D Object Map Building Using Dense Object Models with SIFT-based Recognition Features,” Proceedings of IEEE International Conference on Intelligent Robots and Systems, October 9-15, Beijing, China.
[8] Yamazaki, K., Tomono, M., Tsubouchi, T. and Yuta, S. (2006), “A Grasp Planning for Picking up an Unknown Object for a Mobile Manipulator,” Proceedings of IEEE International Conference on Robotics and Automation, May 15-19, Orlando, Florida, USA.
[9] Grundmann, T., Eidenberger, R., Zoellner, R.D., Xue, Z., Ruehl, Zoellner, J.M., Dillmann, R., Kuehnle, J. and Verl, A. (2008), “Integration of 6D Object Localization and Obstacle Detection for Collision Free Robotic Manipulation,” Proceedings of IEEE International Symposium on System Integration, December 4-4, Nagoya, Japan.
[10] Azad, P., Asfour, T. and Dillmann, R. (2007), “Stereo-Based 6D Object Localization for Grasping with Humanoid Robot Systems,” Proceedings of IEEE International Conference on Intelligent Robots and Systems, Oct 29-Nov 2, San Diego, CA, USA.
[11] Mondragón, I.F., Campoy, P., Martínez, C. and Olivares-Méndez, M.A. (2010), “3D Pose Estimation Based on Planar Object Tracking for UAVs Control,” Proceedings of IEEE International Conference on Robotics and Automation, May 3-7, Anchorage, Alaska, USA.
[12] https://www.epson.com.tw/RobotArm
[13] http://www.rgk-fa.com/
[14] https://www.baslerweb.com/cn/
[15] Benhimane, S. and Malis, E. (2007), “Homography-based 2D Visual Tracking and Servoing,” International Journal of Robotics Research, Vol. 26, No. 7, pp. 661-676.
[16] Viola, P., Jones, M. (2001), “Rapid Object Detection using a Boosted Cascade of Simple Features,” Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition, December 8-14, Kauai, Hawaii, USA.
[17] 張宇宏，「基於顏色與SURF特徵之球場廣告刊版內容辨識與計次系統」（2011），碩士論文，台北科技大學資訊工程研究所，台北市。
[18] Fu, Q. and Wang, H. (2017), “A fast image stitching algorithm based on SURF,” Proceedings of IEEE International Conference on Communications, November 15-16, Kunming, China.
[19] Fischler, M. and Bolles, R. (1981), “Random Sampling Consensus: a Paradigm for Model Ftting with Application to Image Analysis and Automated Cartography,” Commun. Assoc. Comp. Mach., Vol. 24, No. 6, pp. 381-395.
[20] http://en.wikipedia.org/wiki/RANSAC