臺灣博碩士論文加值系統

AOI檢測(Automated Optical Inspection)為運用機器視覺作為檢測的技術，改良傳統人力需配合光學儀器進行檢測的缺點並可增加產品品質的良率。AOI檢測的應用在現代越來越普及，在業界也有許多應用，如印刷檢查、隱形眼鏡檢測、醫療診斷及晶圓檢測等。其主要目的是為減少人力成本、增加產品品質並提升產線效率。隨著科技日新月異的發展及電腦硬體效能的提升，許多產業都朝著自動化的方向發展。
晶圓為最常見的半導體材料，也是IC製造不可或缺的一個元素，其種類非常多樣化，瑕疵的特徵也大不相同，所以在檢測時，如要只用傳統影像演算法去判斷會非常困難，所以本研究主要使用深度學習結合影像辨識之技術，並透過少量樣本，來獲取良好的檢測結果。在深度學習的模型選用上，本研究使用Variational Autoencoder及YOLOv4來做為檢測模型，這兩種架構不僅辨識速度迅速及精準度也很高。檢測流程會先利用YOLOv4將廠商所提供影像中的瑕疵及非瑕疵Pattern抓出，再利用Variational Autoencoder或影像處理的方式來進行驗證，在研究中也有碰上一些問題及困難，之後也會依序去做介紹。本研究的目標為在樣本數較少的情況下也可以利用深度學習進行訓練，並配合電腦視覺的技術自動化即時檢測晶圓瑕疵，取代傳統人力檢查，節省成本增加良率，並設計一套使用者介面可供產線人員操作。

AOI (Automated Optical Inspection) is a technology that uses machine vision to improve the traditional human inspection with optical instruments and increase product quality and yield. The main purpose is to reduce labor costs. The main purpose is to reduce labor costs, increase product quality and improve production line efficiency. As technology advances and computer hardware performance increases, many industries are moving toward automation.
Wafers are the most common semiconductor material and an indispensable element in IC manufacturing. In this study, we use Variational Autoencoder and YOLOv4 as the detection models, which are fast and accurate. The inspection process will first use YOLOv4 to capture the defective and non-defective patterns in the images provided by the manufacturer, and then use Variational Autoencoder or image processing to verify them. The goal of this research is to use deep learning and computer vision technology to develop an automated real-time inspection of wafer defects to replace the traditional human inspection in order to save cost and increase yield, and to design a set of user interface for production line personnel to operate.

摘要 i
ABSTRACT ii
誌謝 iii
目錄 iv
表目錄 vii
圖目錄 viii
第1章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 3
1.3 文獻探討 4
1.4 論文架構 6
第2章 基本理論介紹 7
2.1 卷積神經網路(Convolutional Neural Network) 7
2.1.1 卷積層(Convolutional Layer) 7
2.1.2 池化層(Pooling Layer) 9
2.1.3 平坦層(Flatten Layer) 9
2.1.4 全連接層 (Fully Connecter Layer) 10
2.2 深度學習(Deep Learning) 10
2.2.1 監督式學習(Supervised Learning) 11
2.2.2 非監督式學習(Unsupervised Learning) 12
2.2.3 半監督式學習(Semi-supervised Learning) 12
2.2.4 增強學習(Reinforcement Learning) 12
2.3 分類 (Classification) 12
2.4 定位 13
2.5 權重與偏差(weight and Bias) 13
2.6 激勵函數(Activation Function) 13
2.7 YOLOv4 14
2.3.1 信心值(Confidence) 15
2.8 Autoencoder(AE) 16
2.8.1 Variational Autoencoder（VAE） 17
2.9 二值化 18
2.10 影像相減 18
2.11 模糊化(高斯模糊) 18
第3章 研究方法 19
3.1 瑕疵檢測v1流程介紹 19
3.1.1 訓練集製作 20
3.1.2 VAE訓練 25
3.1.3 訓練結果 25
3.1.4 瑕疵檢測v1流程總結 28
3.2 瑕疵檢測v2流程介紹 29
3.2.1 訓練集製作 29
3.2.2 VAE訓練 30
3.2.3 訓練結果 31
3.2.4 其他模糊化效果比較 33
3.2.5 檢測方式 34
3.2.6 瑕疵檢測v2流程總結 38
3.3 瑕疵檢測v3流程介紹 39
3.3.1 YOLOv4模型訓練 40
3.3.2 VAE訓練 45
3.3.3 訓練結果 45
3.3.4 瑕疵檢測v3流程總結 48
3.4 瑕疵檢測v4流程介紹 49
3.4.1 影像處理 50
3.4.2 瑕疵判斷方式 52
第4章 實驗結果 57
4.1 IC晶圓圖案之自動化瑕疵檢測開發環境 57
4.2 IC晶圓圖案之自動化瑕疵檢測使用者介面 57
4.3 IC晶圓圖案之自動化瑕疵檢測無異常範圍整理 59
4.4 IC晶圓圖案之自動化瑕疵檢測流程重點歸納 60
第5章 結論與未來展望 61
5.1 結論 61
5.5 未來展望 61
參考文獻 62

參考文獻
[1] 遠見雜誌, “晶圓製造、IC 封測市占世界第一，台灣半導體掌握全球科技生命線,” 遠見雜誌, 2021/10/09.
[2] Haiyong Chen , Yue Pang , Qidi Hu and Kun Liu, "Solar Cell Surface Defect Inspection Based on Multispectral Convolutional Neural Network," Journal of Intelligent Manufacturing, vol. 31, no. 2, pp. 453-468, 2020/02/01.
[3] Maoshu Xu ,Haigang Zhang and Jinfeng Yang, "Prohibited Item Detection in Airport X-Ray Security Images via Attention Mechanism Based CNN," in Lecture Notes in Computer Science, 2018/11/02.
[4] H. S. Shon, E. Batbaatar, W. -S. Cho and S. G. Choi, “Unsupervised Pre-Training of Imbalanced Data for Identification of Wafer Map Defect Patterns,” IEEE Access, pp. 52352 - 52363, 4 3 2021.
[5] S. Wang, Z. Zhong, Y. Zhao and L. Zuo, "A Variational Autoencoder Enhanced Deep Learning Model for Wafer Defect Imbalanced Classification," IEEE Transactions on Components, pp. 2055-2060, 12 2021.
[6] J. Cha, S. Oh, D. Kim and J. Jeong, "A Defect Detection Model for Imbalanced Wafer Image Data Using CAE and Xception," in IEEE, Valencia, Spain, 2020/10/19 - 2020/10/22.
[7] A. A. R. M. Abu Ebayyeh, S. Danishvar and A. Mousavi, "An Improved Capsule Network (WaferCaps) for Wafer Bin Map Classification Based on DCGAN Data Upsampling," IEEE, pp. 50-59, 2 2022.
[8] X. Zhang, K. Wang, Y. Wang, Y. Shen and H. Hu, "An improved method of rail health monitoring based on CNN and multiple acoustic emission events," in 2017 IEEE International Instrumentation and Measurement Technology Conference (I2MTC), 2017.
[9] L. Shang, Q. Yang, J. Wang, S. Li, and W. Lei, "Detection of rail surface defects based on CNN image recognition and classification," in in 2018 20th International Conference on Advanced Communication Technology (ICACT), 2018.
[10] Y. Santur, M. Karaköse, and E. Akin, "A new rail inspection method based on deep learning using laser cameras," in in 2017 International Artificial Intelligence and Data Processing Symposium (IDAP), 2017.
[11] G. Tello, O. Y. Al-Jarrah, P. D. Yoo, Y. Al-Hammadi, S. Muhaidat, and U. Lee, "Deep-Structured Machine Learning Model for the Recognition of Mixed-Defect Patterns in Semiconductor Fabrication Processes," in IEEE Transactions on Semiconductor Manufacturing, 2018.
[12] J. Kim, S. Kim, N. Kwon, H. Kang, Y. Kim, and C. Lee, "Deep learning based automatic defect classification in through-silicon Via process: FA: Factory automation," in in 2018 29th Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC), 2018.
[13] P. Kumar, P. Nagar, C. Arora, and A. Gupta, "U-Segnet, "Fully Convolutional Neural Network Based Automated Brain Tissue Segmentation Tool," in in 2018 25th IEEE International Conference on Image Processing (ICIP), 2018.
[14] Z. Huang, M. Fu, K. Ni, H. Sun, and S. Sun, "Recognition of Vehicle-Logo Based on Faster-RCNN," in Signal and Information Processing, Networking and Computers, Springer Singapore, 2019, pp. 75-83.
[15] M. Wang and J. C. P. Cheng, "Development and Improvement of Deep Learning Based Automated Defect Detection for Sewer Pipe Inspection Using Faster R-CNN," in Advanced Computing Strategies for Engineering, 2018, pp. 171-192.
[16] H. Chen, Y. Pang, Q. Hu, and K. Liu, "Solar cell surface defect inspection based on multispectral convolutional neural network," Journal of Intelligent Manufacturing, vol. 31, no. 2, pp. 453-468, 2020.
[17] H. Lin, B. Li, X. Wang, Y. Shu, and S. Niu, "Automated defect inspection of LED chip using deep convolutional neural network," Journal of Intelligent Manufacturing, vol. 30, no. 6, pp. 2525-2534, 2019.
[18] F. Gao, Z. Li, G. Xiao, X. Yuan, and Z. Han, "An online inspection system of surface defects for copper strip based on computer vision," in in 2012 5th International Congress on Image and Signal Processing, 2012.
[19] K. Han, M. Sun, X. Zhou, G. Zhang, H. Dang, and Z. Liu, "A new method in wheel hub surface defect detection: Object detection algorithm based on deep learning," in in 2017 International Conference on Advanced Mechatronic Systems (ICAMechS), 2017.
[20] S. J. Lee, J. Ban, H. Choi, and S. W. Kim, "Localization of slab identification numbers using deep learning," in in 2016 16th International Conference on Control, Automation and Systems (ICCAS), 2016.