基於 YOLO 物件辨識技術之 PCB 多類型瑕疵檢測模型開發

以作者查詢圖書館館藏

、以作者查詢臺灣博碩士

、以作者查詢全國書目

、勘誤回報

、線上人數：63

、訪客IP：18.191.93.95

姓名

曾瀚廣(Han-Kuang Tseng) 查詢紙本館藏

畢業系所

光機電工程研究所

論文名稱

基於 YOLO 物件辨識技術之 PCB 多類型瑕疵檢測模型開發
(Development of PCB Multi-Type Defect Detection Model Based on YOLO Object Recognition Technology)

相關論文

★ MOCVD晶圓表面溫度即時量測系統之開發	★ MOCVD晶圓關鍵參數即時量測系統開發
★ 應用螢光顯微技術強化RDL線路檢測系統	★ 基於人工智慧之PCB瑕疵檢測技術開發
★ 全場相位式表面電漿共振技術	★ 波長調制外差式光柵干涉儀之研究
★ 攝像模組之影像品質評價系統	★ 雷射修整之高速檢測-於修整TFT-LCD SHORTING BAR電路上之應用
★ 光強差動式表面電漿共振感測術之研究	★ 準共光程外差光柵干涉術之研究
★ 波長調制外差散斑干涉術之研究	★ 全場相位式表面電漿共振生醫感測器
★ 利用Pigtailed Laser Diode 光學讀寫頭在角度與位移量測之研究	★ 複合式長行程精密定位平台之研究
★ 紅外波段分光之全像集光器應用	★ 太陽光譜分光器之設計

檔案

[Endnote RIS 格式]

[Bibtex 格式]

[相關文章]

[文章引用]

[完整記錄]

[館藏目錄]

至系統瀏覽論文 ( 永不開放)

摘要(中)

隨著科技日新月異，科技的急速進步驅動硬體技術的飛躍發展，這對電路板的需求不斷攀升，同時也提高了對品質的嚴格要求。深度學習技術因其卓越的應用潛力而備受矚目，不僅在工業界，也在日常生活中發揮了關鍵作用。舉例來說，交通管理領域使用深度學習技術，實現路口科技執法系統，它可以自動偵測紅燈違規或超速行駛，提升道路的安全，同時也提高執法的效率。

目前在電路板印刷相關產業中，印刷電路板（PCB）的良率檢測主要依賴自動光學檢測（AOI）系統和人工檢測。然而AOI系統常常出現缺陷判斷誤差，這導致需要大量人力介入，從而增加了生產成本。為了有效降低PCB檢測的人力成本，本研究提出了一種基於深度學習的檢測技術，用於辨識PCB上的缺陷。我們的目標是建立一個深度學習模型，以高度精確地過濾掉AOI系統標記的「偽缺陷」，從而提升檢測的準確性和效率。

本研究經過一系列嚴謹的測試與評估後，選擇YOLO神經網路作為模型訓練的主架構。近年來YOLO因其在物件偵測領域的卓越性能，已在學術及工業界廣泛應用。本研究將瑕疵視作特定物件，透過深度學習進行細緻的訓練，系統得以高精度地識別並標注瑕疵位置。而模型訓練所用的資料集，則是由合作廠商提供目前AOI系統於產線上所蒐集的瑕疵資料，其中包含了AOI系統錯誤識別的八類瑕疵以及非瑕疵影像資料。

摘要(英)

With the rapid advancement of technology driving the leap forward in hardware techniques, there is an escalating demand for circuit boards, paralleled by increasingly stringent quality requirements. Deep learning technology, recognized for its exceptional potential in applications, plays a pivotal role not only in the industrial sector but also in daily life. For instance, in the field of traffic management, deep learning has been implemented to enable intelligent traffic law enforcement, including technological systems at intersections that automatically detect red light violations or speeding, thereby enhancing road safety and enforcement efficiency.

Currently, in the printed circuit board (PCB) manufacturing industry, the inspection of PCB yield primarily relies on Automated Optical Inspection (AOI) systems and manual checking. However, the AOI systems frequently encounter defect judgment errors, leading to substantial human intervention and thus, increasing production costs. To effectively reduce the labor costs associated with PCB inspection, this study proposes a deep learning-based detection technique to identify defects on PCBs. Our goal is to establish a deep learning model that can accurately filter out the ′pseudo defects′ marked by the AOI systems, thereby increasing the precision and efficiency of inspections.

After a series of rigorous tests and evaluations, this research has chosen the YOLO neural network as the principal framework for model training. YOLO, widely applied in academia and industry for its superior object detection capabilities in recent years, is utilized in this study to treat defects as specific objects. Through meticulous training with deep learning, the system is capable of identifying and marking defect locations with high accuracy. The dataset used for model training is comprised of defect data currently collected by the AOI systems on the production line, provided by our industry partners, including eight types of defects and non-defect image data erroneously identified by the AOI systems.

關鍵字(中)

★ YOLO
★ PCB
★ 瑕疵檢測
★ 深度學習
★ 自動化光學檢測

關鍵字(英)

★ YOLO
★ PCB
★ Defect Detection
★ Deep Learning
★ Automatic Optical Inspection

論文目次

摘要 I
Abstract II
致謝 IV
目錄 V
圖目錄 IX
表目錄 XII
第一章緒論 1
1-1研究背景 1
1-2文獻回顧 3
1-2-1卷積神經網路(CNN) 4
1-2-2 YOLO物件偵測 8
1-3研究動機與目的 11
1-4 論文架構 13
第二章人工智慧與機器學習原理 15
2-1 人工智慧 15
2-1-1弱人工智慧 16
2-1-2強人工智慧 16
2-2 機器學習 17
2-2-1監督式學習 17
2-2-2強化式學習 18
2-2-3非監督式學習 18
2-3 深度學習 19
2-3-1神經網路 19
2-3-2卷積神經網路(CNN) 20
2-3-3 YOLO物件偵測 23
2-4小結 25
第三章實驗架構 26
3-1實驗構想 26
3-2設備規格 27
3-3資料集 28
3-3-1短路特徵 29
3-3-2斷路特徵 30
3-3-3線路凹陷特徵 30
3-3-4突出特徵 31
3-3-5缺口特徵 31
3-3-6壓傷特徵 32
3-3-7異物特徵 32
3-3-8銅顆粒特徵 33
3-4資料標註 33
3-4-1標註軟體 34
3-4-2軟體操作 34
3-4-3標註檔案格式 37
3-5模型架構 38
3-6小節 41
第四章實驗設計 42
4-1實驗流程 42
4-2資料前處理 43
4-2-1資料清洗 43
4-2-2資料平衡 45
4-2-3資料集劃分 50
4-3資料標註方式 51
4-4超參數調整 56
4-4-1 HSV 56
4-4-2角度旋轉 58
4-4-3平移 59
4-4-4上下左右翻轉 60
4-4-5馬賽克 60
4-5小節 61
第五章實驗結果與討論 62
5-1模型評估指標 62
5-1-1混淆矩陣(Confusion Matrix) 62
5-1-2準確率(Accuracy) 64
5-1-3精確率(Precision) 64
5-1-4召回率(Recall) 64
5-1-5誤警率(False Alarm) 65
5-2各類別辨識能力 66
5-3平均精確度 68
5-4 ROC曲線 69
5-5 GradCAM 70
5-6遭遇困難 71
5-6-1瑕疵類別分類錯誤 71
5-6-2特定瑕疵的辨識能力較差 72
5-6-3標註矛盾導致瑕疵類別混淆 74
第六章結論與未來展望 75
6-1結論 75
6-2未來展望 76
參考文獻 77

參考文獻

[1] A. L. Alfonso, "Printed Circuit Boards." U.S. Patent No. 3,033,914. May (1962)
[2] H. Baier, et al, "Method for Automatic Optical Inspection." U.S. Patent No. 4,570,180. Feb (1986)
[3] COGNEX :電子產品業-PCB檢測介紹。2023年10月30日
取自https://reurl.cc/eLOYy7
[4] Z. Linlin, et al, "Convolutional Neural network-based multi-label classification of PCB defects.", IET The Journal of Engineering, Vol 16, pp.1612-1616, (2018)
[5] V. A. Adibhatla, et al, “Detecting Defects in PCB using Deep Learning via Convolution Neural Networks”, 2018 13th International Microsystems, Packaging Assembly and Circuits Technology Conference (IMPACT), pp.202-205, (2018)
[6] Y. S. Deng, A. C. Luo, and M. J. Dai, "Building an Automatic Defect Verification System using Deep Neural Network for PCB Defect Classification." 2018 4th International Conference on Frontiers of Signal Processing (ICFSP), IEEE, (2018)
[7] B. Ghosh, et al, "Defect Classification of Printed Circuit Boards based on Transfer Learning." 2018 IEEE Applied Signal Processing Conference (ASPCON), IEEE, (2018)
[8] J. Deng, et al, "Imagenet: A large-scale hierarchical image database." 2009 IEEE conference on computer vision and pattern recognition, IEEE, (2009)
[9] A. Corovic, et al, "The Real-Time Detection f Traffic Participants Using YOLO Algorithm." 2018 26th Telecommunications Forum (TELFOR), IEEE, (2018)
[10] S. Abbasi, H. Abdi, and A. Ahmadi, “A Face-Mask Detection Approach based on YOLO Applied for a New Collected Dataset.” 2021 26th International Computer Conference, Computer Society of Iran (CSICC), IEEE, (2021).
[11] E. Cengil, A. Çinar, and M. Yildirim. "A Case Study: Cat-Dog Face Detector Based on YOLOv5." 2021 International Conference on Innovation and Intelligence for Informatics, Computing, and Technologies (3ICT), IEEE, (2021)
[12] H. T. Hung, and R. C. Chen. "Pet cat behavior recognition based on YOLO model." 2020 International Symposium on Computer, Consumer and Control (IS3C), IEEE, (2020)
[13] J. McCarthy, et al, "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence.", AI magazine , pp.12-12, (2006)
[14] 創新照顧: AIDA 人流分析解決方案: 人流計數/人流密度/社交距離。2024年1月5號，取自https://reurl.cc/yYM372。
[15] J. C. Flowers, "Strong and Weak AI: Deweyan Considerations." AAAI spring symposium, Towards conscious AI systems. Vol. 2287, No.7, (2019)
[16] A. Kaplan, and M. Haenlein, "Siri, Siri, in my hand: Who’s the fairest in the land ? On the interpretations, illustrations, and implications of artificial intelligence." Business horizons, ScienceDirect, Vol.62, pp.15-25, (2019)
[17] M. I. Jordan, and T. M. Mitchell, "Machine learning: Trends, perspectives, and prospects." ScienceDirect, Vol.349, pp.255-260, (2015)
[18] R. Caruana, and A. Niculescu-Mizil, "An empirical comparison of supervised learning algorithms." Proceedings of the 23rd international conference on Machine learning, pp.161-168, (2006)
[19] L. P. Kaelbling, M. L. Littman, and A. W. Moore, "Reinforcement learning: A survey." Journal of artificial intelligence research, Vol.4, pp.237-285, (1996)
[20] Z. Ghahramani, "Unsupervised learning." Summer school on machine learning, Advanced Lectures on Machine Learning, Pp.72-112, (2003)
[21] Y. LeCun, Y. Bengio, and G. Hinton, "Deep learning." nature Vol.521, pp.436-444, (2015)
[22] A. D. Dongare, R. R. Kharde, and A. D. Kachare, "Introduction to artificial neural network." International Journal of Engineering and Innovative Technology (IJEIT), Vol.2.1, pp.189-194, (2012)
[23] 科學月刊:大腦神經元。2021年7月1日，取自https://reurl.cc/N4R7RQ。
[24] 行銷資料科學:快速反應機制-類神經網路。2019年4月8日，
取自https://reurl.cc/bDEOXy。
[25] Y. LeCun, et al, “Gradient-based learning applied to document recognition.” Proc.IEEE, Vol.86, pp.2278-2324. (1998)
[26] Y. LeCun, and Y. Bengio, “Convolutional networks for images, speech, and time series.” The handbook of brain theory and neural networks, MIT Press, Cambridge, MA, USA, pp.255-258, (1998)
[27] Y. LeCun, et al, “Handwritten digit recognition with a back-propagation network” MIT Press, Cambridge, MA, USA, pp.396-404, (1989)
[28] Ch Tseng :初探卷積神經網路。2017年9月12日，取自https://reurl.cc/v0WZDl。
[29] Yeh James :卷積神經網絡介紹(Convolutional Neural Network)。2017年12月25日，取自https://reurl.cc/RWXMp9。
[30] J. Redmon, et al, "You only look once: Unified, real-time object detection." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp.779-788, (2016)
[31] F. M. Talaat, and H. ZainEldin, "An improved fire detection approach based on YOLO-v8 for smart cities." Neural Computing and Applications, Vol.35, pp.20939-20954, (2023)
[32] J. Terven, and D. Cordova-Esparza. "A comprehensive review of YOLO: From YOLOv1 to YOLOv8 and beyond." arXiv preprint, Computer Vision and Pattern Recognition, pp.1680-1716, (2023)
[33] W. Chen, et al. "YOLO-face: a real-time face detector." The Visual Computer Vol.37, pp.805-813, (2021)
[34] K. Chellapilla, S. Puri, and P. Simard, "High Performance Convolutional Neural Networks for Document Processing." Tenth international workshop on frontiers in handwriting recognition, HAL, (2006)
[35] Tzutalin, Darrenl : LabelImg,。2024年1月8日，取自https://reurl.cc/prMO8r。
[36] G. Van Rossum, "Python Programming Language." USENIX annual technical conference, Vol. 41, No. 1, (2007)
[37] J. Blanchette, and M. Summerfield, "C++ GUI programming with Qt 4." Prentice Hall Professional, (2006)
[38] B. A. Myers, and M. B. Rosson, "Survey on user interface programming." Proceedings of the SIGCHI conference on Human factors in computing systems. (1992)
[39] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with deep convolutional neural networks." In Proceedings of the 25th International Conference on Neural Information Processing Systems, Curran Associates Inc, Red Hook, NY, USA, Vol.1, pp.1097-1105, (2012)
[40] G. Jocher, et al, "ultralytics/yolov5: v6. 2-yolov5 classification models, apple m1, reproducibility, clearml and deci. ai integrations." Zenodo, (2022)
[41] S. Imambi, K. B. Prakash, and G. R. Kanagachidambaresan, "PyTorch." Programming with TensorFlow: Solution for Edge Computing Applications, pp.87-104, (2021)
[42] J. M. Johnson, and T. M. Khoshgoftaar, "Survey on deep learning with class imbalance." Journal of Big Data Vol.6.1, pp.1-54, (2019)
[43] S. Visa, et al, "Confusion matrix-based feature selection." Maics, Vol.710, pp.120-127, (2011)
[44] M. Zhu, "Recall, precision and average precision." Department of Statistics and Actuarial Science, University of Waterloo, Vol.2, pp.6, (2004)
[45] R. R. Selvaraju, et al, "Grad-cam: Visual explanations from deep networks via gradient-based localization." Proceedings of the IEEE international conference on computer vision, (2017)
[46] 黃柏盛，「基於人工智慧之PCB瑕疵檢測技術開發」，國立中央大學，碩士論文，民國111年。
[47] S. Nitish, et al. "Dropout: a simple way to prevent neural networks from overfitting" The journal of machine learning research, Val.15, pp.1929-1958, (2014)

指導教授

李朱育(Lee, Ju-Yi)

審核日期

2024-1-30

推文