為了有效且快速解決生產線上判斷物件的瑕疵與否和解決人工調整參數的問題，本研究開發一套自動調整參數之光學檢測系統，透過影像前處理，去除多餘的資訊，再由一系列的影像處理程序，例如：高斯模糊、影像強化、二值化和K-means等程序，並從中挑選可調整的參數，配合各種參數組合取得影像的量化指標，作為群聚分析之依據，從而判斷物件的瑕疵與否。為了達到自動化調整參數之功能，運用了最佳化方法，其中包括遍歷法、牛頓法與基因演算法，設定目標函數為最小化判斷瑕疵的錯誤數量，藉由反覆迭代，產生新的參數組合，以達到最佳的分類結果。本研究提出一整體準確率，作為評斷最佳參數組合是否是可信賴的數值，進行了一些統計方法，需要各資料點與閥值之間的距離關係，利用兩群中心座標求得閥值方程式與一位於該方程式上的點座標，再進一步計算各資料點至閥值方程式的距離，透過這些距離關係，建立由核密度估計逼近聯合機率密度函數，並由累積機率密度函數計算錯誤的機率，最後可由錯誤的機率取得本研究所提出的條件機率，即為代表該組參數的信賴度。本研究以太陽能板檢測為例，規劃一系列可調整參數的影像處理程序，結果顯示最佳化演算法中以基因演算法運算速度最快且最精準，該組參數信賴度也達到90.9%。

In order to solve the problems, how to judge if the objects have any defects and improve the adjusted parameters manually, effectively and quickly. This research has developed an Automated adjustment of parameters and Automated Optical Inspection (AOI) system. The system removes the unnecessary information by using pre-processing. And then it designs the major image processing steps, such as Gaussian filter, image enhancement, binarization, K-means, etc., to pick some adjustable parameters. Every combination of parameters can acquire image characteristics or factors. According to image characteristics or factors, the system can judge if the objects have any defects. To automatically adjust parameters, this research uses some methods of optimization, such as brute force search, Newton’s method, genetic algorithm (GA). Define the objective function as the minimum number of the errors. Create a new parametric combination by iterating, in order to get the best result of classification. This paper presents an overall accuracy to identify whether a decision of classification is trustable. To compute the conditional probability with statistical methods, it needs to calculate the distance, which is a relationship of data point between the threshold’s function. Threshold’s function is calculated by two centers of the cluster and their middle point. Using the distances to build Kernel Density Estimation (KDE) that is used to approximate the Joint Probability Density Function (JPDF). Cumulative Distribution Function (CDF) is used to calculate the probability of errors that is an important part of computing the conditional probability. The conditional probability is equal to the reliability of parametric combination. In the end, the image classification with an estimation of the conditional probability is used for AOI of a practical inspection of solar cells. The experiment shows that GA is the best algorithm and the reliability of result has reached 90.9%.

[1] Dwight Hooker, Playmate of the Month, in Playboy. 1972.
[2] 簡浩容, 應用實驗設計於 LCD 面板自動化光學瑕疵檢測之影像處理方法選擇, in 工業工程與管理系所. 2015, 國立交通大學.
[3] WJ Zhang, D Li, F Ye, and H Sun. Automatic optical defect inspection and dimension measurement of drill bit. in 2006 International Conference on Mechatronics and Automation. 2006. IEEE.
[4] H Rau and C-H Wu, Automatic optical inspection for detecting defects on printed circuit board inner layers. The International Journal of Advanced Manufacturing Technology, 2005. 25(9-10): p. 940-946.
[5] Chun Ho Wu, Da Zhi Wang, Andrew Ip, Ding Wei Wang, Ching Yuen Chan, and Hong Feng Wang, A particle swarm optimization approach for components placement inspection on printed circuit boards. Journal of Intelligent Manufacturing, 2009. 20(5): p. 535.
[6] "NVIDIA CUDA". Available from: https://developer.nvidia.com/cuda-zone.
[7] Po Ting Lin and Boting Rex Lin. Fuzzy automatic contrast enhancement based on fuzzy C-means clustering in CIELAB color space. in Mechatronic and Embedded Systems and Applications (MESA), 2016 12th IEEE/ASME International Conference on. 2016. IEEE.
[8] 林柏廷, 丁俞文, and 賴韻如, 生物活性配方細胞吸收的對比強化即時影像技術. 2017, 國立臺灣大學系統年輕學者創新性合作計畫聯合補助研究計畫.
[9] Martin P Bendsøe, Topology optimization. 2009: Springer.
[10] 謝士佑, 訊號特徵擷取於銑削加工 Inconel 718 之表面粗糙度預測以及基因遺傳演算法之加工參數最佳化, in 機械工程學系所. 2018, 中興大學. p. 1-67.
[11] Ralph Gross and Vladimir Brajovic. An image preprocessing algorithm for illumination invariant face recognition. in International Conference on Audio-and Video-Based Biometric Person Authentication. 2003. Springer.
[12] Bruce O May, Image preprocessing by modified median filter. 1988, Google Patents.
[13] Ewa Pietka, Arkadiusz Gertych, Sylwia Pospiech, Fei Cao, HK Huang, and Vicente Gilsanz, Computer-assisted bone age assessment: Image preprocessing and epiphyseal/metaphyseal ROI extraction. IEEE transactions on medical imaging, 2001. 20(8): p. 715-729.
[14] 致茂電子股份有限公司, 提供技術支援.
[15] Alexander C Berg and Jitendra Malik. Geometric blur for template matching. in Computer Vision and Pattern Recognition, 2001. CVPR 2001. Proceedings of the 2001 IEEE Computer Society Conference on. 2001. IEEE.
[16] "Mathematical morphology wiki". Available from: https://en.wikipedia.org/wiki/Mathematical_morphology.
[17] T Abatzoglou, A fast maximum likelihood algorithm for frequency estimation of a sinusoid based on Newton's method. IEEE transactions on acoustics, speech, signal processing, 1985. 33(1): p. 77-89.
[18] Beant Kaur and Anil Garg. Mathematical morphological edge detection for remote sensing images. in 2011 3rd International Conference on Electronics Computer Technology. 2011. IEEE.
[19] Lin Wei-Cheng Wayne, Mathematical morphology and its applications on image segmentation, in Dept. of Computer Science Information Engineering. 2000, National Taiwan University.
[20] Glenn Hines, Zia-ur Rahman, Daniel Jobson, and Glenn Woodell, Single-scale retinex using digital signal processors. 2005.
[21] Alexandru Telea, An image inpainting technique based on the fast marching method. Journal of graphics tools, 2004. 9(1): p. 23-34.
[22] 陸韋豪 and 林柏廷, 用在影像處理的高效模糊自動化對比增強法, in 機械工程學系. 2016, 中原大學. p. 1-87.
[23] Tessa K Anderson, Kernel density estimation and K-means clustering to profile road accident hotspots. Accident Analysis and Prevention, 2009. 41(3): p. 359-364.
[24] Anil K Jain, Data clustering: 50 years beyond K-means. Pattern recognition letters, 2010. 31(8): p. 651-666.
[25] Kiri Wagstaff, Claire Cardie, Seth Rogers, and Stefan Schrödl. Constrained k-means clustering with background knowledge. in ICML. 2001.
[26] "Newton's method wiki". Available from: https://en.wikipedia.org/wiki/Newton%27s_method.
[27] Giovanni Dosi and Richard R Nelson, An introduction to evolutionary theories in economics. Journal of evolutionary economics, 1994. 4(3): p. 153-172.
[28] J Maynard Smith, Optimization theory in evolution. Annual Review of Ecology and Systematics, 1978. 9(1): p. 31-56.
[29] Shaker A Zahra and Gerard George, The net-enabled business innovation cycle and the evolution of dynamic capabilities. Information Systems Research, 2002. 13(2): p. 147-150.
[30] Lawrence Davis, Handbook of genetic algorithms. 1991.
[31] Darrell Whitley, A genetic algorithm tutorial. Statistics and Computing, 1994. 4(2): p. 65-85.
[32] 吳明穎 and 張良正, 整合性地下水營運之最佳規劃研究. 2004, National Chiao Tung University.
[33] Emanuel Parzen, On estimation of a probability density function and mode. The annals of mathematical statistics, 1962. 33(3): p. 1065-1076.