本論文將先前本實驗室所開發的輸送帶之物件自動追蹤系統進行了改進，除了從原本的二次組裝升級成三次組裝，同時也把手動校正改成由相機視覺來輔助的自動校正功能，這樣可以提高相機偵測物件位置和角度的精確率。
隨著自動循跡組裝技術的改進，組裝孔和插入物件之間的公差也逐漸縮小至1mm，精確度的提高，則組裝細小乃至精密元件的實驗也越有機會付諸實現。本研究將把被組裝物件延伸至實際產品，以XBOX 360搖桿舉例，其內部有許多零件，若要一次連續組裝完成顯然是不可能的。由於實驗室環境的限制，機械手臂旁只有一架輸送帶，故我們把實驗分成多梯次進行，每次實驗搭配相對應的工作臺和夾爪、夾片等，以模擬工業自動化生產線上連續組裝的場景。
基於組裝物件和被偵測物件不再單一化，所使用的夾爪和夾片也就不能只有一種。針對夾取的不同物件，需設計出能夠順利夾取該物件的夾片或夾具，同時也要考慮夾爪本身的開合度，以防在組裝時與被組裝物件發生誤觸，造成物體偏移。而由於物件的不規則形狀，我們設計可以用來偵測且可固定被組裝物件的治具，輪廓既簡單又實用。再搭配換爪機構進行快速換爪，便能組裝不同的零件。上述系統，可適用各種規則與不規則形狀的產品，加速組裝效率，具有產業價值。

This thesis proposes to improve the automatic object tracking system on conveyor developed by our laboratory. In addition to upgrading from a two-time assembly to a three-time assembly implementation, we also changed the manual calibration to the auto-calibrated function supported by computer vision to improve the accuracy of camera detection of object’s position and angle.
With the improvement of the technique of auto-tracking assembly system, the tolerance between the assembly hole and inserting object can also be reduced to 1mm, by which we have better chance to conduct the assembly tasks involving precise operations. This research will extend to implement on a commercial product, the XBOX 360 controller. There are many components inside, and it is impossible to perform continuous assembly tasks in one time with our facility. Therefore, we divide the assembly tasks into several separate operations, each part equipped with a corresponding platform and a proper gripper in order to demonstrate a scene which is a continuous assembly on the industrial automatic production line.
Based on no simplification of assembly and detected object, we cannot use only one kind of gripper and need to design a gripper which is able to precisely grasp the object under consideration. Meanwhile, we also apply a proper gap tolerance of the gripper in order to prevent accidental contacts with the assembled object when during grasping, and causing the offset of the object. For tackling irregular shapes of objects, we design a jig which can be detected and fix the assembled object. The edge of the jig is simple and practical. We combined with the gripper changing mechanism to change grippers so that they can assemble different items. This system described above is suitable for the products which contain many kinds of regular or irregular shapes and it can speed up the assembly efficiency with a high industrial value.

[1] 林冠甫，「基於影像伺服之輸送帶上物件自動追蹤系統」，碩士論文，台灣
科技大學，2016。
[2] 鍾國亮，影像處理與電腦視覺第五版，東華書局，第120-126頁，2012。
[3] 莊靖煒，「客製化多功能塗膠組裝生產線」，碩士論文，虎尾科技大學，
2016。
[4] Richard’s blog 攝像頭校正 Camera calibration
http://goo.gl/J90CGr
[5] Heping Chen, George Zhang, et al. "Flexible Assembly Automation Using Industrial Robots" 2008 IEEE International Conference on Technologies for Practical Robot Applications, IEEE, 2008.
[6] Fang Xu,Shuai Wang,Bangyu Li. " Industrial robot base assembly based on improved Hough transform of circle detection algorithm" Proceeding of the 11th World Congress on Intelligent Control and Automation, IEEE, 2014.
[7] Jaeyoul Lee , Ingyu Park , Kwangwoo Jeon , Sanghun Lee. "An Experimental Evaluation for the Precision Assembly of Rotational Sensor According to the Changes of Fit Tolerance" 2013 10th International Conference on Ubiquitous Robots and Ambient Intelligence (URAI), IEEE,2013.
[8] M.Peña, I. López, R. Osorio. "Invariant Object Recognition Robot Vision System for Assembly" Electronics, Robotics and Automotive Mechanics Conference (CERMA'06),Vol.1, IEEE,2006.
[9] Songgi LEE, Xin JIANG, Koyu ABE, Satoko ABIKO, Atsushi KONNO and Masaru UCHIYAMA. "A Human Assistant Robot System for Handling Heavy Mechanical Parts in Assembly Lines" 2011 IEEE/SICE International Symposium on System Integration (SII), IEEE,2011.
[10] Young-LouI Kim I , Hee-Chan Song, Jae-Bok Song. "Force Control based Jigless Assembly Strategy of a Unit Box Using Dual-Arm and Friction" IEEE ISR 2013, IEEE,2013.
[11] Feifei.Zhao,Hao Gu,Cheng Li,Chanjuan Chen. "Accuracy Analysis for Robotized Assembly System" 2017 IEEE International Conference on Robotics and Biomimetics (ROBIO), IEEE,2017.
[12] Jim Takamats,Koichi Oganara,Hiroshi Kitnurat,Katsushi Ikench. "Correcting Observation Errors for Assembly Task Recognition" IEEE/RSJ International Conference on Intelligent Robots and Systems, IEEE,2002.