軸承為現代機械及車輛之主要支撐形式，被廣泛運用，包括國防、通訊衛星及航空等，其中軸承精度及安裝好壞標準嚴謹，所以對軸承生產品質需予更精準地檢測。目前市面上，大多仍依靠人眼辨識表面瑕疵及使用多樣儀器量測軸承幾何尺寸，但人眼檢測具有下列缺失：1.每個人對瑕疵標準不一。2.長時間工作下，容易疲憊，導致漏檢或誤檢狀況。3.需花費大量人力及耗費時間。
本研究旨在開發一套自動化軸承瑕疵檢測系統，透過機器視覺與硬體機構結合，達到自動檢測軸承幾何尺寸瑕疵及表面瑕疵。本系統主要對軸承四種部分進行檢測，包括端面、圓周面、內部及圓周面型態，其中又分為幾何尺寸瑕疵檢測及表面瑕疵檢測。首先對工業相機進行比例校正，計算實際尺寸與像素間比例，並輸入軸承檢測之標準規格，讓系統作為篩選依據。接著透過最小平方圓方法，得到軸承上下端面真圓度及圓尺寸；再使用座標轉換及正規化灰階變異數檢測法，計算端面表面瑕疵。利用旋轉台及影像拼接，擷取軸承圓周面及內部影像，使用線段偵測得到圓周面高度尺寸；透過傅立葉轉換，計算頻率圖上平均能量值曲率最大之最佳半徑，並去除中心點及最佳半徑外頻率元素，將影像反傅立葉轉換，將影像二值化分割，計算圓周面表面瑕疵；使用正規化灰階變異數檢測法，對內部表面瑕疵進行偵測。透過背光燈源打光圓周面，藉由線段及亮度個數，能分辨加工痕有無或大小是否一致。利用自動化軸承瑕疵檢測系統，能將物件特徵標準一致化，全天不停歇檢測，提高生產效率，降低人力成本，減少產品不良率、誤判率及漏檢率。

Bearings are the main form of support for modern machinery and vehicles, and are widely used, including national defense, communications satellites and aviation. Among them, bearing precision and installation standards are rigorous, so the quality of bearing production needs to be more accurately detected. At present, most of the factories still rely on the human eye to identify the surface defects and use a variety of instruments to measure the bearing geometry, but the human eye detection has the following defects: 1. Each person has different standards. 2. Under long hours of work, it is easy to be exhausted, resulting in missed or missed detection conditions. 3. It takes a lot of manpower and time.
This study aims to develop an automated bearing defect inspection system that combines machine vision with hardware to automatically detect bearing geometry and surface flaws. The system mainly tests the four parts of the bearing, including the end face, the circumferential surface, the inner and the circumferential surface type, and detect geometric size and surface defect. First, the industrial camera is scaled, the actual size and pixel ratio are calculated, and the standard specifications of bearing inspection are input to make the system a screening basis. Then, through the least square circle method, the roundness and the circle size of the upper and lower end faces of the bearing can be obtained; and the surface flaw is calculated by using the coordinate conversion and the normalized gray-scale variation number detection method. Using the rotating table and image Stitching, the bearing circumferential surface and internal image are captured, and the circumferential surface height dimension is obtained by using the line segment detection; the Fourier transform is used to calculate the optimal radius of the average energy value curvature on the frequency map, deleting the center point and outside the radius is converted by inverse Fourier transform. The image is binarized and divided to calculate the surface flaws of the circumferential surface. The internal surface flaws are detected using normalized gray-scale variation detection method. By illuminating the circumferential surface through the backlight, the number of lines and brightness can be used to distinguish whether the processing marks are uniform or the size is the same. The automated bearing defect detection system can consistently standardize the object characteristics, non-stop testing throughout the day, improve production efficiency, reduce labor costs, and reduce product defect rate, false positive rate and missed detection rate.

論文審定書 i
誌謝 ii
中文摘要 iii
英文摘要 iv
目 錄 vi
圖 次 ix
表 次 xi
第一章 緒論 1
1.1研究背景與目的 1
1.2相關研究 4
第二章 系統架構 8
2.1硬體架構 8
2.1.1工業相機 9
2.1.2外同軸光源 10
2.2軟體架構 11
2.2.1資料管理 12
2.2.2影像校正 13
2.2.3帳號管理 14
2.2.4檢測結果報告 16
第三章 研究方法 18
3.1第一&三檢測站 18
3.1.1輪廓偵測 19
3.1.2圓形逼近法 20
3.1.3座標轉換 22
3.1.4正規化灰階變異數檢測法 24
3.2第二檢測站 26
3.2.1傅立葉轉換 28
3.2.2內部檢測 30
3.3第四檢測站 31
3.3.1 Sobel邊緣偵測 32
3.3.2線性迴歸(Linear Regression) 33
3.3.3灰階值計算 35
第四章 實驗結果 36
第五章 結論與未來展望 41
參考文獻 42

[1]W. Lord, and D. Oswald, “Leakage Field Methods of Defect Detection,” International Journal of Nondestructive Testing, Vol. 4, Issue 12, pp. 249-274, 1972.
[2]J. R. Leslie, and W. J. Cheesman, “An Ultrasonic Method of Studying Deterioration and Cracking in Concrete Structures,” American Concrete Institute Journal, Vol. 46, Issue 9, pp. 17-36, 1949.
[3]Keiji Tsukada, Takuya Tomioka, Shunki Wakabayashi, Kenji Sakai, and Toshihiko Kiwa, “Magnetic Detection of Steel Corrosion at a Buried Position Near the Ground Level Using a Magnetic Resistance Sensor,” IEEE Transactions on Magnetics, pp. 1-4, 30 May 2018
[4]M. Adzal, and S. Udpa, “Advanced Signal Processing of Magnetic Flux Leakage Data Obtained from Seamless Steel Pipeline,” NDT&E International, Vol. 35, pp. 449-457, 2002.
[5]Milan Sonka, Vaclav Hlavac, and Roger Boyle, “Image Processing, Analysis, and Machine Vision,” Cengage Learning, pp. 582-782, 2014
[6]Xiaoqin Liu, Xing Wu, and Chang Liu, “A Comparison of Acoustic Emission and Vibration on Bearing Fault Detection,” 2011 International Conference on Transportation, Mechanical, and Electrical Engineering, Changchun, China, 16-18 Dec. 2011.
[7]J.B Catlin, “The Use of Ultrasonic Diagnostic Techniques to Detect Rolling Element Bearing Defects,” Proceedings of Machinery Vibration Monitoring and Analysis Meeting, Vibration Institute, pp. 123–130, Houston, TX, 19-21 April 1983.
[8]R. Ito, and T. Azuma, “Development of Non-destructive Grinding Burn Detection Technology,” Journal of JTEKT Engineering, 2015.
[9]Maya. V. Lakha, and K. Chennakeshava Reddy, “Image Processing Application for Non Destructive Testing,” IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI-2017), Chennai, India, 21-22 Sept. 2017.
[10]W. Xian, Y. Zhang, Z. Tu, and E.L. Hall, “Automatic Visual Inspection of the Surface Appearance Defects of Bearing Roller,” IEEE International Conference on Robotics and Automation, Vol. 3, pp. 1490-1494,13-18 May 1990.
[11]李韋辰，“應用機器視覺於 PU-密封元件之瑕疵檢測”，中華大學，機械與航太工程研究所，碩士論文，2006
[12]S.M. Thomas, and Y.T. Chan, “A Simple Approach for the Estimation of Circular Arc Center and its Radius,” Computer Vision, Graphics, and Image Processing, Vol. 45, Issue 3, pp. 362-370, 1989.
[13]Tao He, Jinshan Zhou, Na Dai, Zhihui Chen, and Li Du, “The Defects Identify Algorithm and the Automated Detecting System for the Screw Thread,” 2010 International Conference on Electrical and Control Engineering, Wuhan, China 25-27 June 2010.
[14]Nam Kyu Kwon, Jong Seok Lee, and PooGyeon Park, “Mura Region Detection by Using 2D FFT with Exponential Kernel for Black Resin-Coated Steel,” International Journal of Information and Electronics Engineering, Vol. 2, No. 6, November 2012.
[15]T.C. Chen, and K.L. Chung. “An Efficient Randomized Algorithm for Detecting Circles,” Computer Vision and Image Understanding, pp. 172-191, 2001.
[16]N. Kanopoulos, N. Vasanthavada, and R.L. Baker, “Design of an Image Edge Detection Filter Using the Sobel Operator,” IEEE Journal of Solid-State Circuits, Vol. 23, Issue 2, pp. 358-367, Apr. 1988.