基於影像處理及FPN模型的PCB瑕疵檢測｜國立臺灣科技大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	范順強 Shun-Qiang Fan
論文名稱：	基於影像處理及FPN模型的PCB瑕疵檢測 PCB defect detection based on image processing and FPN model
指導教授：	楊振雄 Chen-Hsiung Yang
口試委員:	郭永麟 Yong-Lin Kuo 陳金聖 Chin-Sheng Chen 吳常熙 Chang-Shi Wu
學位類別：	碩士 Master
系所名稱：	工程學院 - 自動化及控制研究所 Graduate Institute of Automation and Control
論文出版年：	2020
畢業學年度：	108
語文別：	英文
論文頁數：	85
中文關鍵詞：	印刷電路板、瑕疵檢測、深度學習、FPN模型、影像融合
外文關鍵詞：	Printed circuit board, Defect detection, Deep Learning, FPN, Image fusion
相關次數：	點閱：354 下載：0
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

印刷電路板的瑕疵檢測是電子設備生產過程中不可缺少的一道過程，目前傳統的方法是基於範本比對的影像處理方法。隨著印刷電路板的功能越來越豐富，同樣尺寸的印刷電路板上的電路佈局也愈加複雜，這使得傳統的演算法在準確率和效能上愈顯不足。為了應對當前印刷電路板的發展趨勢，本論文提出了一種結合影像處理及深度學習技術的瑕疵檢測方法，旨在提升針對印刷電路板瑕疵的檢測精度及效能。論文主要工作內容如下：
(1)印刷電路板中的瑕疵種類豐富且特徵各異，其中短路及斷路這兩類瑕疵具有特徵不易識別的特點，在檢測過程中往往容易出現誤判及漏判。針對於此我們採用一種基於明度疊加的影像融合方法，增強了短路及斷路這兩類瑕疵的特徵，改善這兩類瑕疵檢測率的同時不影響其他類型瑕疵的檢測。
(2)將深度學習中基於區域推薦的特徵金字塔網絡(Feature Pyramid Networks， FPN)應用於印刷電路板各類瑕疵的檢測任務。在FPN框架中，我們採用ResNet-101網絡提取目標特徵，再通過特徵金字塔融合算法將不同層次的特徵圖進行融合，最後將融合的特徵圖送入RPN網絡及分類網絡，實現了對各類瑕疵的檢測及類別判定。
我們使用經過影像融合的增強資料集訓練FPN模型，經過實驗結果顯示，針對印刷電路板各類瑕疵檢測的mAP(Mean Average Precision)達到97.29%。

Defect detection of PCB (Printed Circuit Board) is an indispensable process in the production process of electronic equipment. The current traditional method is an image processing method based on template comparison. As the functions of PCB become more and more abundant, the circuit layout of PCB of the same size becomes more and more complicated, which makes the traditional algorithms more and more insufficient in accuracy and performance. In order to cope with the current development trend of PCB, this thesis proposes a defect detection method combining image processing and deep learning technology, aiming to improve the detection accuracy and efficiency of PCB defects. The main content of the thesis is as follows:
(1) The types of defects in the PCB are rich and have different characteristics. Among them, the two types of defects, short circuit and open circuit, have characteristics that are not easy to identify, and it is often prone to misjudgment and missed judgment during the detection process. In view of this, we adopt an image fusion method based on lightness superimposition, which enhances the characteristics of two types of defects, short circuit and open circuit, and improves the detection rate of these two types of defects without affecting the detection of other types of defects.
(2) The FPN (Feature Pyramid Networks) based on region proposal in deep learning is applied to the detection of various defects in PCB. In the FPN framework, we use the ResNet-101 network to extract object features, and then use the feature pyramid fusion algorithm to fuse feature maps at different levels. Finally, the fusion feature map is sent to the RPN network and the classification network to achieve defect detection and category determination.
We use the enhanced dataset after image fusion to train the FPN model. The experimental results show that the mAP (Mean Average Precision) of various defects detection for PCB reaches 97.29%.

誌謝 ................................................................................................................................... I
摘要 ................................................................................................................................... II
Abstract ...........................................................................................................................III
CONTENTS .................................................................................................................... IV
List of Figure ................................................................................................................. VI
List of Table ....................................................................................................................VIII
Chapter 1 Introduction...............................................................................................1
1 Introduction.............................................................................................................1
2 Motivation and Purpose.....................................................................................2
3 Literature Review...................................................................................................2
3.1 Traditional PCB detection methods............................................................2
3.2 Detection method based on deep learning............................................4
4 Outline.......................................................................................................................6
Chapter 2 Dataset Enhancement and Image Fusion .....................................8
1 Original PCB Defect Dataset..............................................................................8
2 PCB Defect Dataset Enhancement.................................................................13
2.1 Random Crop Algorithm................................................................................13
2.2 Expansion of the PCB Defects Dataset .....................................................14
3 Image Fusion Algorithm for PCB Feature Enhancement ......................16
Chapter 3 Deep Learning Framework..................................................................20
1 CNN Feature Extractor........................................................................................20
1.1 Convolutional Neural Network....................................................................20
1.2 ResNet ..................................................................................................................28
2 Region Proposal Network.................................................................................33
3 Classification Network........................................................................................39
3.1 ROI Pooling .........................................................................................................40
3.2 Fully Connected Network...............................................................................41
3.3 Non-Maximum Suppression.........................................................................42
4 Feature Pyramid Network..................................................................................45
Chapter 4 Experimental Results and Discussion..............................................49
1 Experimental Environment................................................................................49
2 PCB Dataset.............................................................................................................52
3 Evaluation Standard ............................................................................................53
4 FPN Model Results ..............................................................................................56
4.1 FPN Model Training Process.........................................................................56
4.2 Hyper Parameter Setting................................................................................57
4.3 Analysis of FPN Model Training Process .................................................59
4.4 Analysis of FPN Model Experiment Results.............................................63
4.5 FPN Model Test Results...................................................................................65
Chapter 5 ........................................................................................................................69
1 Conclusion................................................................................................................69
2 Future Work.............................................................................................................70
Reference.........................................................................................................................71
                                

[1] J. A. Magera, G. J. Dunn, inventors; Motorola Solutions Inc, assignee, “Printed circuit board,” United States patent US 7, 459, 202, 2008.
[2] L. I. Zheng-Ming, L. I. Hong, and Jun Sun. “Detection of PCB based on digital image processing,” Instrument Technique & Sensor, vol. 61, no. 8, pp. 87–89, 2012.
[3] H. S. Cho, J. G. Yoo, J. S. Kim, S. H. Kim, inventors; Samsung ElectroMechanics Co Ltd, assignee, “Printed circuit board,” United States patent US 8, 159, 824, 2012.
[4] R. T. Chin, C. A. Harlow, “Automated visual inspection: A survey,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-4, no. 6, pp. 557–573, Nov. 1982, doi:10.1109/TPAMI.1982.4767309.
[5] M. Moganti, F. Ercal, “Automatic PCB inspection systems,” Potentials, vol. 14, no. 3, pp. 6–10, 1995.
[6] W. Y. Wu, M. Jiun, et al. “Automated inspection of printed circuit boards through machine vision,” Computers in Industry, vol. 28, no. 2, pp.103–111, 1996.
[7] W. Sterling, "Automatic Non-reference Inspection of Printed Wiring Boards," IEEE Pattern Recognition & Image Processing, New York, pp. 458-460, 1979.
[8] S. Kumar, Y. Iwahori, and M. K. Bhuyan. Proceedings of International Conference on Computer Vision and Image Processing, vol. 459, 2017.
[9] F. Sabath and H. Garbe, “The Analysis of PCB Layout Using a Hybrid MoM-MTL Method,” IEEE EMC Symposium, vol. 2, pp. 774-779, 1998.
[10] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, et al., “Backpropagation applied to handwritten zip code recognition,” Neural Computation, vol. 1, no. 4, pp. 541-551, 1989.
[11] Y. Lecun, L. Bottou, Y. Bengio and P. Haffner, “Gradient-based learning applied to document recognition,” Proceedings of the IEEE, vol. 86, no. 11, pp. 2278-2324, Nov. 1998.
[12] A. Krizhevsky, I. Sutskever, G. E. Hinton. “ImageNet classification with deep convolutional neural network,” International Conference on Neural Information Processing Systems, Curran Associates Inc., 2012.
[13] M. D. Zeiler, R. Fergus. “Visualizing and Understanding Convolutional Networks,” European Conference on Computer Vision, vol. 8689, pp. 818-833, 2014.
[14] K. Simonyan and A. Zisserman. “Very deep convolutional networks for large-scale image recognition,” International Conference on Learning Representations, 2015.
[15] K. He, X. Zhang, S. Ren and J. Sun, "Deep Residual Learning for Image Recognition," IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, pp. 770-778, Nov. 2016.
[16] Y. Jia, E. Shelhamer, et al. “Caffe: Convolutional Architecture for Fast Feature Embedding,” Proceeding of the ACM International Conference on Multimedia, ACM, Orlando, pp. 675-678, 2014.
[17] M. Abadi, A. Agarwal, P. Barham, et al. “Tensorflow: Large-scale machine learning on heterogeneous distributed systems,” 12th USENIX Symposium on Operating Systems Design and implementation, Savannah, GA-USA, vol. 16, pp. 265-283, 2016.
[18] J. Redmon, S. Divvala, R, Girshick, et al. “You only look once: unified, real-time object detection,” IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 2016.
[19] T. Chen, M. Li, Y. Li, et al. “Mxnet: A flexible and efficient machine learning library for heterogeneous distributed systems,” Workshop on Machine Learning Systems, Neural Information Processing Systems, Montreal, 2015.
[20] R. Girshick, J. Donahue, T. Darrell, et al. “Rich feature hierarchies for accurate object detection and semantic segmentation,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, DC, USA, 580-587, 2014.
[21] K. He, X. Zhang, S. Ren and J. Sun, “Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 9, pp. 1904-1916, 2015.
[22] R. Girshick, “Fast R-CNN,” Proceedings of IEEE International Conference on Computer Vision. Los Alamitos; Washington; Tokyo, pp. 1440-1448, 2015.
[23] S. Ren, K. He, R. Girshick, et al. “Faster R-CNN:towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, 2017.
[24] T. Y. Lin, P. Dollar, R. Girshick, et al. “Feature pyramid networks for object detection,”Computer Vision and Pattern Recognition, Hawaii, America, pp. 936–944, July 2017.
[25] W. Liu, D. Anguelov, D. Erhan, et al. “SSD: single shot multibox detector,” European Conference on Computer Vision, pp. 21-37, 2016.
[26] J. Redmon, A. Farhadi. “YOL09000: Better, Faster, Stronger,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, New York, pp. 6517-6525, 2017.
[27] J. Redmon, A. Farhadi. “YOLOv3: an incremental improvement,” arXiv preprint arXiv: 1804.02767, 2018.
[28] W. B. Huang and P. Wei, “A PCB dataset for defects detection and classification,” IEEE Conference on Computer Vision and Pattern Recognition, vol. 14, no. 8, pp. 1-9, 2018.
[29] L. Demidova, I. KLYUEVA, “SVM classification: optimization with the SMOTE algorithm for the class imbalance problem,” Proceedings of IEEE Conference on mbedded Computing, Pisca NJ, pp. 1-4, 2017.
[30] S. Kumar, Y. Iwahori, and M. K. Bhuyan. “PCB Defect Classification Using Logical Combination of Segmented Copper and Non-copper Part,” Proceedings of International Conference on Computer Vision and Image Processing, vol.459, pp. 523-532, 2017.
[31] P. Vincent, H. Larochelle, Y. Bengio, et al. “Extracting and composing robust features with denoising autoencoders,” International Conference on Machine Learning, pp. 1096-1103, 2008.
[32] D. Scherer, A. Miiller, S. Behnke, “Evaluation of pooling operations in convolutional architectures for object recognition,” Artificial Neural Networks-ICANN, pp. 92-101, 2010.
[33] J. R. Uijlings, K. D. Sande, T. Gevers, et al. “Selective search for object recognition,”International Journal of Computer Vision, vol. 104, no. 2, pp. 154-171, 2013.
[34] R. S. Sutton, A. G. Barto, “Evaluative Feedback,” in Reinforcement Learning: An Introduction , MITP, pp.25-49, 1998.
[35] S. X. Zhang and R. L. Qiang, “An Algorithm for Computing the Union, Intersection and Difference of Two Polygons,” Proceedings of International Conference on Computer Research and Development, Kuala Lumpur, pp. 344-348, 2010.
[36] A. Neubeck and L. V. Gool, “Efficient Non-Maximum Suppression,” 18th International Conference on Pattern Recognition, Hong Kong, pp. 850-855, 2006.
[37] OpenCV[EB/OL]. http://zh.wikipedia.org/wiki/OpenCV, Apr. 2014.
[38] C. Metz. “Google Just Open Sourced TensorFlow, Its Artificial Intelligence Engine,” Wired. Retrieved November 10, 2015.
[39] Stehman, V. Stephen. “Selecting and interpreting measures of thematic classification accuracy,” Remote Sensing of Environment. vol. 62, no. 1, pp. 77–89, 1997.
[40] C. Szegedy et al., “Going deeper with convolutions,” IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, pp. 1-9, 2015

全文公開日期 2025/07/08 (校內網路)
全文公開日期本全文未授權公開 (校外網路)
全文公開日期本全文未授權公開 (國家圖書館：臺灣博碩士論文系統)

簡易檢索 / 詳目顯示

相關論文