本研究目的在開發出一套，以三條線雷射與一台攝影機建構架設於機械手臂法蘭面之自動循線系統，此系統可用於即時偵測三維曲面上之非特定軌跡線，並全自動引導機械手臂執行循線運動，藉以執行特定加工任務，使機械手臂可免於執行任務之事前教導作業。
自動循線系統部分，首先攝影機進行校正完後，予以線雷射與攝影機之關係校正，其校正過程將使單條線雷射投射至水平作用面，並於不同深度位置，針對雷射線與黑白校正板進行取像，分別對三條線雷射進行此動作，即可於後續分析導出三條線雷射與攝影機之深度關係式。透過攝影機與機械手臂之關係校正，即可將自動循線系統計算之三維資訊，用於引導機械手臂執行相關任務。
為驗證此系統的精確度，本研究進行三項實驗，分別為三轉軸之估算分析、空間作用偏移量之分析、和實際引導手臂作業之偏移量分析等。
轉角之估算分析將會使自動循線系統，於單一轉軸之不同角度，對固定平面之特定軌跡線與黑白校正板進行取像，分別對三轉軸進行此動作，即可分析出此系統於三轉軸之角度估算誤差。空間作用偏移量之分析，即針對空間已知一條直線，進行自動循跡，並將計算之三維資訊，與已知的三維資訊作分析，以得三維空間系統估算之偏移量。
最終設置三維空間上之軌跡線，並利用自動循線系統，引導機械手臂執行特定任務，於本實驗設置為塗膠任務。當機械手臂被引導完成塗膠任務後，其被作用樣本將與被作用前樣本進行比對，即可得系統於實際引導手臂作業之偏移量。

The purpose of this study is to develop a 6-axis Robot Autonomous Line Tracking and Following System with three line lasers and a camera. The system can autonomously follow and track a 3D line and thereby conduct some processing by robot arm without time-consuming teaching tack.
In the autonomous line tracking system, we must conduct some calibration. First, we need to perform camera calibration. Second, let the camera plane be parallel to the working space, which is the same plane with the check board. Then project a line laser on the working space and take images in different depths of camera and working space for calibrating the relationship of the line laser and the camera. By the relationship between the line laser and the camera, the calibration performed on each line laser can get three equations to calculate the depth of the camera and the working area. By calibration of the camera and the tool coordinate, the vision system can connect to the robot arm to conduct some processing on the line tracking.
Some experiments are conducted to prove the system accuracy, on the three-axis rotation calculate errors, the work point deviation in 3D space, and the gluing result deviation.
In the three-axis rotation calculation error analysis, only the work on one axis is independently conducted in each experiment. To take the image on the checkboard in difference angle, we can get the system slope in the 3D space. By comparing two values from the checkboard and the vision system, we can get the three-axis rotation calculation error.
On the work point deviation analysis, we need to generate the 3D model of a line in 3D space. Using the tracking system to scan the line to get the work points. After comparing the value of the 3D model and the real work points, we can calculate the deviation between the work points and the track line.
In the gluing experiment, we can compare the gluing result and the track line sample to get the deviation of the gluing accuracy.

[1]林崇賢與江耀宗，機器人應用實務，全華科技圖書，1989。
[2]許郁文、曹茹蘋、蔣智揚、與謝嘉文譯，機器人科技，馥林文化，2014。
[3]龔仲華與龔曉雯，親手操作控制及校調工業機器人，佳魁數位，2017。
[4] Jae Byung Park, Seung Hun Lee, and Il Jae Lee. "Precise 3D lug pose detection sensor for automatic robot welding using a structured-light vision system." Sensors, Vol. 9, No. 9, 2009, pp. 7550-7565.
[5]李冠誼，應用結構光投影之三維量測與循跡技術，碩士論文，國立臺灣科技 大學機械工程研究所，台北(2014)。
[6]Scansonic MI GmbH, U.S. Patent No. 2011/0290780 (01 December, 2011)
[7]Gary Bradski, and Adrian Kaehler. "Learning OpenCV: Computer vision with the OpenCV library." O'Reilly Media, Inc., 2008, pp. 375-378.
[8]蔡高岳，機械系統設計，高立圖書，2004。
[9]Peter Corke. Robotics, vision and control: fundamental algorithms in MATLAB. Vol. 73. Springer, 2011, pp. 455-456.
[10]OpenCV 2.4.13.4 Documentation, Camera Calibration with OpenCV. https://goo.gl/sdp8hJ
[11]聯安達國際有限公司，技術支持。
https://goo.gl/BCr92G