跳到主要內容

臺灣博碩士論文加值系統

(44.192.48.196) 您好!臺灣時間:2024/06/14 18:36
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

: 
twitterline
研究生:吳嘉恩
研究生(外文):WU,JIA-EN
論文名稱:利用生成對抗網路生成自動光學檢查之瑕疵影像資料
論文名稱(外文):Generation of Automated Optical Inspection Images by Generative Adversarial Network
指導教授:朱學亭朱學亭引用關係
指導教授(外文):CHU,HSUEH-TING
口試委員:王玲玲陳朝欽朱學亭
口試委員(外文):WANG,LING-LINGCHEN,CHAUR-CHINCHU,HSUEH-TING
口試日期:2019-06-28
學位類別:碩士
校院名稱:亞洲大學
系所名稱:資訊工程學系
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2019
畢業學年度:107
語文別:中文
論文頁數:29
中文關鍵詞:深度學習生成對抗網路
外文關鍵詞:Deep LearningGenerative Adversarial NetworkAutomated Optical Inspection
相關次數:
  • 被引用被引用:1
  • 點閱點閱:306
  • 評分評分:
  • 下載下載:2
  • 收藏至我的研究室書目清單書目收藏:0
自動光學檢查(Automated Optical Inspection,簡稱AOI),是應用機器視覺技術來偵測生產線上製造的物件是否有嚴重缺陷,提升產品良率。近幾年,深度學習模型應用在AOI上有許多成功的案例,可以強化原先非深度學習模型演算法的AOI儀器所檢測的正確性。傳統的深度學習依賴大量的測試資料,才能達到最好的結果。然而,生產線上製造的物件可能種類繁多。對於每一項產品,都要收集到足夠的瑕疵資料以訓練深度學習模型,有時並非那麼容易。因此,若能快速增加瑕疵的影像資料可以加速AOI深度學習模型訓練以及提高正確性。
近年來利用生成對抗網路 (Generative Adversarial Network; 簡稱GAN ) 來合成影像引起了很多研究的關注。生成對抗網路包括生成網路模組(generative network)和判別網路模組(discriminative network)。本論文研究利用生成對抗網路來產生各種AOI瑕疵資料。在我們的實驗中,透過生成對抗網路產生的瑕疵資料,提升AOI深度學習模型的正確性2.1%。實驗結果說明生成對抗網路可以用來產出瑕疵影像以提升自動光學檢查的正確性。

Automated Optical Inspection (AOI) is the application of machine vision technology to detect whether the objects manufactured on the production line have serious defects and improve product yield. In recent years, there have been many successful cases in the application of deep learning models on AOI, which can enhance the correctness of AOI instruments that were originally used in non-deep learning model algorithms. Traditional deep learning relies on a large amount of test data to achieve the best results. However, there are a wide variety of objects that can be manufactured on the production line. For each product, it is not always easy to collect enough data to train a deep learning model. Therefore, if you can quickly increase the image data of defect, you can accelerate the training of AOI deep learning model and improve the correctness.
In recent years, the use of the Generative Adversarial Network (GAN) to synthesize images has attracted much attention. A Generative Adversarial Network includes a generative network and a discriminative network. This thesis studies the use of Generative Adversarial Network to generate various AOI data. In our experiments, the correctness of the AOI deep learning model was improved by 2.1% by generating defective images with the proposed Generative Adversarial Network. The experimental results show that the Generative Adversarial Network can be used to generate defect images to improve the accuracy of automated optical inspection.

摘要 I
ABSTRACT II
圖目錄 IV
表目錄 I
第一章緒論 1
1.1研究背景與動機 1
1.2研究目的 1
1.3論文架構 2
第二章背景知識與文獻探討 4
2.1自動化光學檢測瑕疵分類 4
2.2深度學習 7
2.2.1 ResNet模型 7
2.2.2 VGG模型 8
2.3 生成對抗網絡 9
第三章研究設計 16
3.1研究流程 16
3.2資料來源 17
3.3圖片處理與電腦設備 22
3.4 利用深度學習對自動化光學檢測瑕疵分類 23
3.5 利用生成對抗網路生成瑕疵圖像 23
3.6 實驗結果 26
第五章結論與未來目標 27
5.1 結論 27
5.2 未來目標 27
參考文獻 28


圖目錄
圖 1: 論文架構 2
圖 2:自動光學檢測流程圖 5
圖 3:VGG模型圖 9
圖 4:生成對抗網路流程 11
圖 5:GAN所生成的手寫圖片 12
圖 6: 生成對抗網路論文成長趨勢 13
圖 7:線性整流函數(Rectified Linear Unit, ReLU) 14
圖 8:DCGAN 生成器的模型圖 14
圖 9:DCGAN 生成利用celebA數據及所生成的人臉 15
圖 10:研究流程圖 16
圖 11:AIdea首頁 17
圖 12:AIdea AOI議題 18
圖 13:AOI Label 0 19
圖 14:AOI Label 1 19
圖 15:AOI Label 2 20
圖 16:AOI Label 3 20
圖 17:AOI Label 4 21
圖 18:AOI Label 5 21
圖 19:影像裁切 22
圖 20:label 1比較,左邊為原始圖片,右邊為生成出來的圖片 24
圖 21:label 2比較,左邊為原始圖片,右邊為生成出來的圖片 24
圖 22:label 3比較,左邊為原始圖片,右邊為生成出來的圖片 25
圖 23:label 4比較,左邊為原始圖片,右邊為生成出來的圖片 25
圖 24:label 5比較,左邊為原始圖片,右邊為生成出來的圖片 26



表目錄
表 1:訓練資料集 18
表 2 : 各模型準確度比較 23
表 3:各模型再加上生成圖片準確度比較 26


參考文獻
[1]張智崴, "碳奈米纖維生產缺陷之自動光學檢測分類," 碩士, 電子工程研究所, 中原大學, 桃園縣, 2018.
[2]林佑任, "電路板零件及焊點之自動光學檢測系統," 碩士, 電子工程系, 南臺科技大學, 台南市, 2018.
[3]張綺純, "絲餅瑕疵的自動光學檢測," 碩士, 資訊學院資訊學程, 國立交通大學, 新竹市, 2017.
[4]H. G. d. Silva, T. G. Amaral, and O. P. Dias, "Automatic optical inspection for detecting defective solders on printed circuit boards," in IECON 2010 - 36th Annual Conference on IEEE Industrial Electronics Society, 2010, pp. 1087-1091.
[5]李秉鴻, "機器學習於自動光學檢測之應用," 碩士, 資訊工程學系碩士班, 義守大學, 高雄市, 2010.
[6]劉芝妤, "利用提升決策樹建構液晶面板自動光學檢測分類模型," 碩士, 工業工程與管理系所, 國立交通大學, 新竹市, 2016.
[7]徐碩駿, "基於深度學習之自動光學檢測演算法," 碩士, 資訊工程系, 國立臺灣科技大學, 台北市, 2018.
[8]D. Lim, Y. Kim, and T. Park, "SMD Classification for Automated Optical Inspection Machine Using Convolution Neural Network," in 2019 Third IEEE International Conference on Robotic Computing (IRC), 2019, pp. 395-398.
[9]A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with deep convolutional neural networks," presented at the Proceedings of the 25th International Conference on Neural Information Processing Systems - Volume 1, Lake Tahoe, Nevada, 2012.
[10]K. Simonyan and A. Zisserman, "Very Deep Convolutional Networks for Large-Scale Image Recognition," arXiv e-prints, Accessed on: September 01, 2014Available: https://ui.adsabs.harvard.edu/abs/2014arXiv1409.1556S
[11]K. He, X. Zhang, S. Ren, and J. Sun, "Deep Residual Learning for Image Recognition," arXiv e-prints, Accessed on: December 01, 2015Available: https://ui.adsabs.harvard.edu/abs/2015arXiv151203385H
[12]A. Graves, M. Liwicki, S. Fernández, R. Bertolami, H. Bunke, and J. Schmidhuber, "A Novel Connectionist System for Unconstrained Handwriting Recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 31, no. 5, pp. 855-868, 2009.
[13]S. Hochreiter, #252, and r. Schmidhuber, "Long Short-Term Memory," Neural Comput., vol. 9, no. 8, pp. 1735-1780, 1997.
[14]J. Deng, W. Dong, R. Socher, L. Li, L. Kai, and F.-F. Li, "ImageNet: A large-scale hierarchical image database," in 2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009, pp. 248-255.
[15]I. J. Goodfellow et al., "Generative Adversarial Networks," arXiv e-prints, Accessed on: June 01, 2014Available: https://ui.adsabs.harvard.edu/abs/2014arXiv1406.2661G
[16]GAN ZOO. Available: https://github.com/hindupuravinash/the-gan-zoo
[17]A. Radford, L. Metz, and S. Chintala, "Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks," arXiv e-prints, Accessed on: November 01, 2015Available: https://ui.adsabs.harvard.edu/abs/2015arXiv151106434R
[18]A. Brock, J. Donahue, and K. Simonyan, "Large Scale GAN Training for High Fidelity Natural Image Synthesis," arXiv e-prints, Accessed on: September 01, 2018Available: https://ui.adsabs.harvard.edu/abs/2018arXiv180911096B
[19]T. Karras, S. Laine, and T. Aila, "A Style-Based Generator Architecture for Generative Adversarial Networks," arXiv e-prints, Accessed on: December 01, 2018Available: https://ui.adsabs.harvard.edu/abs/2018arXiv181204948K


QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top
無相關期刊