因應時代的進展，產品品質檢測為現代化工業製造重要環節之一。生產過程中，若能即時發現產品瑕疵，既能保證產品品質，也能有效提高生產效率。因此本研究對於印刷瑕疵檢測，建立了一生成對抗式網路模型之異常檢測系統，進行印刷瑕疵檢測模型研究。
本研究主要分為三個步驟，第一步驟蒐集影像資料並對其進行印刷區域裁切，利用模板匹配技術結合影像旋轉，定位出印刷範圍區域並裁切，最後進行黑色填充調整圖像尺寸；第二步驟將進行影像擴增，藉由改變亮度、色溫、對比度以及影像旋轉角度等，對影像進行資料擴增；第三步驟進行異常檢測模型訓練，藉由本研究使用之重建式模型的特殊性，將影像進行重建，並比對重建誤差及影像特徵向量，進行異常分數計算，藉此檢測瑕疵。
本研究於擺放條件隨機及半隨機(自然光源)條件下蒐集之印刷隨身碟外殼影像，藉由影像定位及資料擴增技術，結合少量樣本學習概念，僅利用少量(10張)正常影像進行模型訓練，實現印刷瑕疵檢測。在 300張正常以及300張異常之印刷隨身碟外殼測試影像，檢測準確率達到98.17%、精確度率到97.07% 以及召回率達到99.33%。

In response to the advancements of the times, product quality inspection has become a crucial aspect of modern industrial manufacturing. Detecting product defects in real-time during the production process not only ensures product quality but also effectively enhances production efficiency. Therefore, this research establishes an anomaly detection system using a generative adversarial network (GAN) model for the detection of printing defects.
This research is primarily divided into three steps. The first step involves collecting image data and performing print area cropping. Using template matching technology combined with image rotation, the print area is located and cropped, followed by black filling to adjust image size. The second step involves image augmentation, where image data is augmented by altering brightness, color temperature, contrast, and rotation angle. The third step involves training the anomaly detection model. Utilizing the specificity of the reconstruction model used in this research, the images are reconstructed, and the reconstruction errors and image feature vectors are compared to calculate anomaly scores for defect detection.
This research focuses on detecting printing defects on USB drive cases collected under random and semi-random (natural light) placement conditions. By employing image localization and data augmentation techniques, along with the concept of few-shot learning, the model is trained using only a small number (10) of normal images to achieve printing defect detection. Testing on 300 normal and 300 defective printed USB drive case images, the detection accuracy reached 98.17%, precision rate was 97.07%, and recall rate was 99.33%.

摘要 I
Abstract II
致謝 IV
目錄 VII
圖目錄 XI
表目錄 XV
第一章 緒論 1
1.1 前言 1
1.2 研究動機 2
第二章 文獻回顧 4
2.1 影像處理 4
2.1.1 影像去噪 4
2.1.2 影像旋轉 5
2.1.3 資料擴增 6
2.1.4 少量樣本學習 7
2.2 機器學習 8
2.2.1 類神經網路 9
2.2.2 深度神經網路 10
2.2.3 卷積神經網路 11
2.2.4 生成對抗式網路 12
2.3 異常檢測 18
2.3.1 生成對抗式網路的異常檢測技術 19
2.3.2 其它異常檢測技術 23
第三章 研究方法 28
3.1 建構資料集 29
3.1.1 拍攝環境 29
3.1.2 正常與異常資料 31
3.1.3 異常種類 32
3.2 影像前處理 33
3.2.1 印刷範圍提取 33
3.2.2 影像尺寸調整 35
3.3 資料擴增 36
3.3.1 影像旋轉 36
3.3.2 顏色變換 37
3.4 異常檢測模型 38
3.4.1 生成器與鑑別器網路架構 38
3.4.2 損失函數 40
3.5 異常評估指標與性能 41
3.5.1 異常檢測方式 41
3.5.2 異常檢測模型性能評估 44
3.6 消融實驗 46
3.6.1 印刷範圍提取實驗 46
3.6.2 影像數量 47
3.6.3 影像方向 47
3.6.4 影像擴增、數量以及種類 47
3.6.5 參數選擇 47
第四章 結果與討論 49
4.1 資料與印刷範圍提取結果 49
4.2 異常檢測模型結果與分析 51
4.2.1 實驗結果 52
4.2.2 結果驗證、分析與可視化 54
4.3 消融實驗結果 59
4.3.1 印刷範圍提取實驗結果 59
4.3.2 模型連接層之重建測試 62
4.3.3 資料擴增實驗 63
4.3.4 不同種類印刷測試 65
4.3.5 重建適應值 67
第五章 結論與未來工作 69
5.1 結論 69
5.2 研究貢獻 69
5.3 未來展望 71
參考文獻 73

[1] I. Goodfellow et al., "Generative adversarial nets," Advances in neural information processing systems, vol. 27, pp. 139-144, 2014.
[2] H. H. K. Tin, "Removal of noise by median filtering in image processing," 6th Parallel Soft Comput.(PSC 2011), pp. 1-3, 2011.
[3] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," presented at the Proceedings of the IEEE conference on computer vision and pattern recognition, 2016.
[4] C. Shorten and T. M. Khoshgoftaar, "A survey on image data augmentation for deep learning," Journal of big data, vol. 6, no. 1, pp. 1-48, 2019.
[5] Y. Wang, Q. Yao, J. T. Kwok, and L. M. Ni, "Generalizing from a few examples: A survey on few-shot learning," ACM computing surveys (csur), vol. 53, no. 3, pp. 1-34, 2020.
[6] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "Imagenet classification with deep convolutional neural networks," Advances in neural information processing systems, vol. 25, pp. 1097-1105, 2012.
[7] 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, 1998.
[8] A. Radford, L. Metz, and S. Chintala, "Unsupervised representation learning with deep convolutional generative adversarial networks," arXiv preprint arXiv:1511.06434, 2015.
[9] P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros, "Image-to-image translation with conditional adversarial networks," presented at the Proceedings of the IEEE conference on computer vision and pattern recognition, 2017.
[10] M. Mirza and S. Osindero, "Conditional generative adversarial nets," arXiv preprint arXiv:1411.1784, 2014.
[11] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18, pp. 234-241, 2015.
[12] T. Karras, S. Laine, and T. Aila, "A style-based generator architecture for generative adversarial networks," Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4401-4410, 2019.
[13] T. Karras, T. Aila, S. Laine, and J. Lehtinen, "Progressive growing of gans for improved quality, stability, and variation," arXiv preprint arXiv:1710.10196, 2017.
[14] T. Schlegl, P. Seeböck, S. M. Waldstein, U. Schmidt-Erfurth, and G. Langs, "Unsupervised anomaly detection with generative adversarial networks to guide marker discovery," International conference on information processing in medical imaging, pp. 146-157, 2017.
[15] Z. Zhao, B. Li, R. Dong, and P. Zhao, "A surface defect detection method based on positive samples," PRICAI 2018: Trends in Artificial Intelligence: 15th Pacific Rim International Conference on Artificial Intelligence, Nanjing, China, August 28–31, 2018, Proceedings, Part II 15, pp. 473-481, 2018.
[16] T. Ojala, M. Pietikäinen, and D. Harwood, "A comparative study of texture measures with classification based on featured distributions," Pattern recognition, vol. 29, no. 1, pp. 51-59, 1996.
[17] S. Akcay, A. Atapour-Abarghouei, and T. P. Breckon, "Ganomaly: Semi-supervised anomaly detection via adversarial training," Computer Vision–ACCV 2018: 14th Asian Conference on Computer Vision, Perth, Australia, December 2–6, 2018, Revised Selected Papers, Part III 14, pp. 622-637, 2019.
[18] S. Akcay, A. Atapour-Abarghouei, and T. P. Breckon, "Skip-ganomaly: Skip connected and adversarially trained encoder-decoder anomaly detection," 2019 International Joint Conference on Neural Networks (IJCNN), pp. 1-8, 2019.
[19] T. Schlegl, P. Seeböck, S. M. Waldstein, G. Langs, and U. Schmidt-Erfurth, "f-AnoGAN: Fast unsupervised anomaly detection with generative adversarial networks," Medical image analysis, vol. 54, pp. 30-44, 2019.
[20] M. Arjovsky, S. Chintala, and L. Bottou, "Wasserstein GAN," arXiv preprint arXiv:1701.07875, 2017.
[21] I. Gulrajani, F. Ahmed, M. Arjovsky, V. Dumoulin, and A. C. Courville, "Improved training of wasserstein gans," Advances in neural information processing systems, vol. 30, pp. 5767-5777, 2017.
[22] J. An and S. Cho, "Variational autoencoder based anomaly detection using reconstruction probability," Special lecture on IE, vol. 2, no. 1, pp. 1-18, 2015.
[23] P. Bergmann, M. Fauser, D. Sattlegger, and C. Steger, "Uninformed students: Student-teacher anomaly detection with discriminative latent embeddings," Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4183-4192, 2020.
[24] S. Wang, L. Wu, L. Cui, and Y. Shen, "Glancing at the patch: Anomaly localization with global and local feature comparison," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 254-263, 2021.
[25] J. Wyatt, A. Leach, S. M. Schmon, and C. G. Willcocks, "Anoddpm: Anomaly detection with denoising diffusion probabilistic models using simplex noise," Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 650-656, 2022.
[26] J. Ho, A. Jain, and P. Abbeel, "Denoising diffusion probabilistic models," Advances in neural information processing systems, vol. 33, pp. 6840-6851, 2020.
[27] 趙宇涵, "針印自動化異常檢測之研究," 碩士論文, 動力機械工程學系, 國立清華大學, 新竹市, 2023. [Online]. Available: https://hdl.handle.net/11296/mu7d5b