現今自動化檢測流程大都透過機構或振動盤的方式讓料桶中的零件轉成特定方位，再利用工業相機進行定位或人工排列的方式將零件放置至特定位置，以利機械手臂將零件夾至檢測區域，但此方式需針對不同零件來進行調整且浪費人力，為克服此問題，本研究開發「自動化隨機拾取與瑕疵檢測系統」，以減少手臂夾取前的前置處理工作，並搭配表面影像檢測，讓自動化檢測流程更加彈性。
本研究主要分成三大部分：零件隨機拾取、檢測視角決定及零件瑕疵檢測，其中“零件隨機拾取”是利用3D點資料演算法分析料桶內所有零件的點雲資料，尋找適當的零件及適合機械手臂吸取該零件的位置，並使用EPSON機械手臂將零件以特定方位放置於檢測區域；“檢測視角決定”是以零件整體外觀來判定該零件適合以哪些基本視角（上、下、左、右、前、後）進行檢測；“零件瑕疵檢測”是使用零件外觀的2D影像來偵測刮痕及凹洞，以決定產品是否為良品。
本論文除了說明隨機拾取及瑕疵檢測所使用到的演算法外，亦針對多個不同幾何特性的產品來測試隨機拾取的實用性，而瑕疵檢測因須針對零件特性開發不同演算法，故僅針對華接頭進行測試。

The typical process of automatic inspection for defective parts usually utilizes some mechanism to obtain specific orientations of the parts in order for a robotic arm to grab the parts to an inspection area. Such a process is time consuming and is hard to fit the need of different parts. In order to reduce the pre-processing time of robot grasping, this thesis studies the issue of automatic bin picking for random placement of parts. Automatic defect inspection based on surface images of the parts is also undertaken to make the inspection work more flexible.
This study consists of the following three research issues: automatic random bin picking of parts, automatic determination of inspection views and automatic defect inspection. The first issue “automatic random bin picking of parts” discusses the development of several algorithms using the 3D data points obtained by scanning the whole parts placed in a tank. The algorithms are used to automatically find the orientation of each part in the tank. With the orientations of all parts, the system automatically decides the part to be picked and it’s appropriate grasping position for a robot arm. After that, an EPSON robot arm is used to move the part to an inspection zone. The second issue “automatic determination of inspection views” is the use of the part’s exterior appearance to determine which views (top view, bottom view, front view, back view, left view and right view) are proper for the part to be fully inspected. The last issue “automatic defect inspection” focuses on detecting the scratches and dents on a part’s exterior appearance based on the 2D image taken from the part.
In random bin picking, in addition to proposing solution algorithms, this study tests a number of parts with different geometric characteristics to prove the stability and applicability of the developed system. While for defect inspection, different algorithms should be developed for different parts, thus this study will only take thread pipe as an example for implementation.

[1] Fredik, K. (2007), Methods for Real-Time Bin-Picking using 2D Vision, Master Thesis, School of Computer Science and Communication, KTH Royal Institute of Technology, Stockholm, Sweden
[2] Besl, P.J. and McKay, N.D. (1992), “A Method for Registration of 3-D Shapes,” IEEE Trans. on Pattern Analysis and Machine Intelligence, Vol. 14, No. 2.pp. 239-256.
[3] Rusinkiewicz, S. and Marc L.(2001), “Efficient Variants of the ICP Algorithm,” Proceedings Third International Conference on 3-D Digital Imaging and Modeling, May 28-June 1, Quebec, Canada.
[4] Zhang, Z.Y. (1994), “Iterative point matching for registration of free-form curves and surfaces,” International Journal of Computer Vision. Springer, Vol. 13, No. 12, pp. 119-152
[5] Pomerleau F. , Colas F. and Siegwart R. (2015), “ A Review of Point Cloud Registration Algorithms for Mobile Robotics,” Foundations and Trends in Robotics, Vol. 4, No. 1, pp. 1-104
[6] Low, K.L. (2004), “Linear Least-Squares Optimization for Point-toPlane ICP Surface Registration,’’ Technical Report, TR04-04
Department of Computer Science, University of North Carolina, Chapel Hill, NC, USA.
[7] Rusu, R.B., Blodow, N. and Beetz, M. (2008), “Towards 3D point cloud based object maps for household environment,” Robotics and Autonomous Systems Journal, Vol. 56, No. 11, pp. 927-941.
[8] Rusu, R.B., Blodow, N. and Beetz, M. (2009), “Fast point feature histograms (FPFH) for 3D registration,” IEEE International Conference on Robotics and Automation, May 12-17 , Kobe, Japan.
[9] Fischler, M.A. and Bolles, R.C. (1981), “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,” Communications of the ACM, Vol. 24, No. 11, pp. 381-395.
[10] 唐培文，「基於結構光之自動隨機物件夾取系統」（2015），碩士論文，台灣科技大學機械工程系研究所，台北市。
[11] Akhloufi, M.A. (2012), “A 2D-3D Hybrid Vision System for Robotic Manipulation of Randomly Oriented Objects,” International Journal of Mechanical and Mechatronics Engineering, Vol. 6, No. 11, pp. 1307-6892
[12] Furtado1, L.F.F., Trabasso1, L.G., Villani1 E. and Francisco A.
(2012), “Temporal Filter Applied to Image Sequences Acquired by an Industrial Robot to Detect Defects in Large Aluminum Surfaces Areas,” International Conference on Mechatronics, Dec 5-7, Prague, Czech Republic.
[13] Fan, X.J., Wang, X.L. and Xiao, Y.F. (2014), “A Combined 2D-3D Vision System for Automatic Robot Picking,” International Conference on Advanced Mechatronic Systems, Aug 10-12, Kumamoto, Japan.
[14] Alam, M.A., Ali, M.M., Musaddeque, A.S., Nawaj S. and Rahaman, M.A. (2014), “An Algorithm to Detect and Identify Defects of Industrial Pipes Using Image Processing,” International Conference on Software, Knowledge, Information Management and Applications, Dec 18-20, Dhaka, Bangladesh.
[15] Acharya, T. and Ray, A.K. (2005), Image Processing: Principles And Applications, John Wiley & Son, 2005. Inc., Los Angeles, CA, USA, pp. 77.
[16] https://www8.hp.com/us/en/campaign/3Dscanner/overview.html
[17] https://www.baslerweb.com/
[18] https://www.epson.com.tw/RobotArm
[19] http://www.rgk-fa.com/
[20] Rublee, E., Rabaud, V., Kurt, K. and Gary, B. (2011), “ORB: An
efficient alternative to SIFT or SURF,” IEEE international conference, Nov 6-13, Barcelona, Spain, pp. 2564-2571
[21] Benhimane, S. and Malis, E. (2004), “Homography-based 2D Visual Tracking and servoing,” International Journal of Robotics Research, Vol. 26, No. 7, pp. 661-676.
[22] 黃健堯，「逆向工程之點雲配對研究」（2014），碩士論文，雲林科技大學機械工程系研究所，雲林縣。
[23] 周宣宏，「棧板箱型物件全自動上下載系統」（2017），碩士論
文，台灣科技大學機械工程系研究所，台北市。