臺灣博碩士論文加值系統

製鞋業為台灣重要的製造業之一，製鞋的過程非常繁複需要經過兩百多道工序。
皮革為製鞋業常用的材料，目前鞋業的皮革表面瑕疵品檢方式主要是以人工目視確認皮革上是否有瑕疵，過程耗時且長時間工作下容易使人眼疲勞，容易造成檢測上的疏失。本研究提出兩階段的視覺檢測方法，此方法結合影像處理與深度學習方法，在本研究設計的光學環境使用影像處理技術檢測皮革表面較明顯的瑕疵，搭配深度學習(YOLOv3 Model)來輔助判定不明顯瑕疵與良品。
根據本研究提出的檢測流程針對100張不良品影像與良品影像進行檢測，以驗證檢測流程，第一階段的DIP檢測可檢測出58%瑕疵，第二階段的YOLO檢測可檢測出40%瑕疵，總檢出率達98%。透過本研究提出的兩階段檢測方法，有助於減少製鞋業目檢作業所需的人力與成本。

The shoes and footwear industry is one of the most important manufacturing fields in Taiwan. It is considered a traditional and complicated handicraft process with more than 200 operations. Leather is a commonly used material and the leather surface inspection method of the footwear industry mainly relies on manual visual inspection, the process is time consuming and long-term work is easy to cause eye fatigue, which is easy to cause negligence in detection. In this study, we proposed a two-phased visual inspection approach which integrates digital image processing and deep learning method In the first phase, we developed a digital image process (DIP) algorithm to detect obvious defects. The deep learning model( Yolo3) used to detect the unobvious defects.
According to the detection process proposed in this study, used another 100 pieces of defective leathers and non-defective leather to verify the detection process. The proposed DIP-based method in the first-phase screened out the 58% of large defect, and then the trained Yolo3 model further detected 40% defect, and reached a 98% inspect rate in total. By means of the two-phased inspection design helps to reduce the manpower and cost required for visual inspection in the footwear industry.

摘 要 i
ABSTRACT ii
誌 謝 iv
目 錄 v
表目錄 vii
圖目錄 ix
第一章 緒論 1
1.1研究背景與動機 1
1.2研究目的 2
1.3研究範圍與限制 2
1.4研究架構 3
第二章 文獻探討 4
2.1皮革瑕疵檢測相關文獻回顧 4
2.2影像處理演算法回顧 6
2.2.1二值化 7
2.2.2形態學 8
2.2.3低通濾波器 10
2.3機器學習演算法回顧 11
2.3.1主成份分析 11
2.3.2 自動編碼器 12
2.3.3 支持向量機 13
2.3.4倒傳遞類神經網路 14
2.4深度學習演算法回顧 16
2.4.1卷積神經網路 16
2.4.2 物件偵測神經網路 19
2.4.3YOLO系列演算法 20
2.4.4 深度學習模型評估 25
第三章 研究方法 26
3.1硬體架構 26
3.2軟體架構 27
3.3調整皮革檢測環境 28
3.4 DIP檢測 32
3.3.1斷開運算 33
3.3.2黑帽運算 33
3.3.3 二值化 34
摘 要 i
ABSTRACT ii
誌 謝 iv
目 錄 v
表目錄 vii
圖目錄 ix
第一章 緒論 1
1.1研究背景與動機 1
1.2研究目的 2
1.3研究範圍與限制 2
1.4研究架構 3
第二章 文獻探討 4
2.1皮革瑕疵檢測相關文獻回顧 4
2.2影像處理演算法回顧 6
2.2.1二值化 7
2.2.2形態學 8
2.2.3低通濾波器 10
2.3機器學習演算法回顧 11
2.3.1主成份分析 11
2.3.2 自動編碼器 12
2.3.3 支持向量機 13
2.3.4倒傳遞類神經網路 14
2.4深度學習演算法回顧 16
2.4.1卷積神經網路 16
2.4.2 物件偵測神經網路 19
2.4.3YOLO系列演算法 20
2.4.4 深度學習模型評估 25
第三章 研究方法 26
3.1硬體架構 26
3.2軟體架構 27
3.3調整皮革檢測環境 28
3.4 DIP檢測 32
3.3.1斷開運算 33
3.3.2黑帽運算 33
3.3.3 二值化 34
3.3.4高斯濾波 34
3.5 YOLO 檢測 35
第四章 實驗成果與分析 38
4.1 調整皮革檢測環境實驗 38
4.1.1 BPN迴歸模型訓練實驗結果 38
4.1.2 SVM迴歸模型訓練實驗結果 41
4.2 YOLO檢測實驗 44
4.2.1樣本及實驗平台 44
4.2.2 YOLO模型訓練實驗結果 45
4.2.3 YOLO模型分類正確率之驗證 48
4.3皮革瑕疵檢測系統實驗 50
第五章 討論 51
5.1 兩階段檢測 51
5.2 系統召回率 51
第六章 結論與未來研究方向 52
6.1結論 52
6.2未來研究方向 52
參考文獻 53

1.Song, K.; Yan, Y. A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects. Appl. Surf. Sci,vol.285,2013,pp858–864.
2.Wu, Y.; Qin, Y.; Wang, Z.; Jia, L,A UAV-based visual inspection method for rail surface defects. Appl. Sci,vol.8,2018,pp10-28.
3.Chondronasios,A.; Popov, I.; Jordanov, I. Feature selection for surface defect classification of extruded aluminum profiles. Int. J. Adv. Manuf. Technol., vol.83, 2016,pp 33–41
4.Cen, Y.G.; Zhao, R.Z.; Cen, L.H.; Cui, L.H.; Miao, Z.J.;Wei, Z. Defect inspection for TFT-LCD images based on the low-rank matrix reconstruction. Neurocomputing, vol.149, 2015,pp1206–1215.
5.De Araújo, S.A.; Pessota, J.H.; Kim, H.Y. Beans quality inspection using correlation-based granulometry. Eng. Appl. Artif. Intell.,vol.40, 2015,pp84–94.
6.Gibert, X.; Patel, V.M.; Chellappa, R. Deep multitask learning for railway track inspection. IEEE Trans. Intell. Transp. Syst. vol.18, 2017,pp153–164.
7.Nian Cai , Guandong Cen, Jixiu Wu, Feiyang Li, Han Wang, and Xindu Chen, SMT Solder Joint Inspecti.on via a Novel Cascaded Convolutional Neural Network. IEEE Vol.8, NO. 4, 2018.,pp99-106
8.Lin, H.; Li, B.; Wang, X.; Shu, Y.; Niu, S. Automated defect inspection of LED chip using deep convolutional neural network. J. Intell. Manuf. , 2018,pp1–10.
9.Parag Kohli,Ms. Shalvi Garg, Leather Quality Estimation Using an Automated Machine Vision System. IOSR-JECE,vol.6,2013,pp44-47.
10.Li Jian, Han Wei, He Bin, Research on Inspection and Classification of Leather Surface Defects Based on Neural network and Decision Tree. ICCDA,China,2010,pp381-384.
11.Roberto Viana, Ricardo B. Rodrigues, Marco A. Alvarez,Hemerson Pistori, SVM with Stochastic Parameter Selection for Bovine Leather Defect Classification. , PSIVT Santiago, Chile,2007,pp.600-612.
12.Wen Wang,,Automatic Visual Inspection for Leather Manufacture.Key Engineering Materials Vols.326-328,2006,pp 469-472 .
13.Lidiya Georgieva,Kaloyan Krastev,Identification of Surface Leather Defect, CompSysTech,Varna,Bulgaria,2005,pp303-307.
14.Choonjong Kwak, Jos´e A. Ventura,,Karim Tofang-Sazic,Automated defect inspection and classification of leather fabric. Intelligent Data Analysis,Vol.5,2001,pp355–370.
15.D.-M. Tsai ,S.-K. Wu, Automated Surface Inspection Using Gabor Filters, Vol.16,2000, pp 474–482.
16.K. Hoang , W. Wen, A. Nachimuthu, X.L. Jiang, Achieving automation in leather surface inspection,ELSEVER,Vol.34,October 1997, pp43-54.
17.G.Vinodhini, RM.Chandrasekaran, Sentiment Classification Using Principal Component Analysis Based Neural Network Model. IEEE, Chennai,India,2014.
18.Md. Omar Faruqe ; Md. Al Mehedi Hasan,Face recognition using PCA and SVM. IEEE, Hong Kong,China,2009.
19.Francisco J. Pulgar,Francisco Charte,Antonio J. Rivera,Mar J. del Jesus, AEkNN: An AutoEncoder kNN-based classier with built-in dimensionality reduction, International Journal of Computational Intelligence Systems,vol.12,2019,pp436-452.
20.Yao Ju, Jun Guo, Shuchun Liu, A Deep Learning Method Combined Sparse Autoencoder with SVM. IEEE, Xi'an,China,2015,pp257-260.
21.Zhang Xuewu, Xu Lizhong, Ding Yanqiong, Fan Xinnan, Automated visual inspection of surface defects based on compound moment invariants and support vector machine. HIGH TECHNOLOGY LETTERS,vol.18,2012,pp.26-32.
22.Smriti H Bhandari,A Simple Approach to Surface Defect Detection. IEEE, Kharagpur,
INDIA,2008.
 
23.Yi Murphey , Jianjun Shi,An Intelligent Real-time Vision System for Surface Defect Detection. IEEE, Cambridge,UK,2004.
24.Ruifang Ye, Ming Chang, Chia-Sheng Pan, Cheng An Chiang, and Jacque Lynn Gabayno ,High-resolution optical inspection system for fast detection and classification of surface defects. International Journal of Optomechatronics,vol.12,2018,pp1-10.
25.Gnanaprakash.V , Sathishkumar.N , Finney Daniel Shadrach S,Back Propagation Neural Network for Defect Detection of Woven Fabrics,International Journal of Computer Applications,vol.86,2014,pp11-14.
26.Li Yi,Guangyao Li, and Mingming Jiang,An End-to-End Steel Strip Surface Defects Recognition System Based on Convolutional Neural Networks, steel research,vol.88, 2017,pp1-12.
27.Shiyang Zhou, Youping Chen, Dailin Zhang, Jingming Xie, Yunfei Zhou,
Classification of Surface Defects on Steel Sheet Using Convolutional Neural,Materials and technology,vol.51,2017,pp123-131.
28.Limei Song,Xinyao Li,Yangang Yang,Xinjun Zhu,Qinghua Guo,Huaidong Yang,Detection of Micro-Defects on Metal Screw Surfaces Based on Deep Convolutional Neural Networks,Sensors,2018,vol.18,pp1-14.
29.IanJ. Goodfellow, Jean Pouget-Abadie_, Mehdi Mirza, Bing Xu, David Warde-Farley,Sherjil Ozairy, Aaron Courville, Yoshua Bengio, Generative Adversarial Nets,NIPS,2014.
30.Maayan Frid-Adar, Idit Diamant, Eyal Klang, Michal Amitai, Jacob Goldberger, GAN-based Synthetic Medical Image Augmentation for increased CNN Performance in Liver Lesion Classification. ELSEVIER Vol.321,2018,pp321-331.

31.Nasibeh Esmaeilishahmirzadi1, Hamidreza Mortezapour, A novel method for enhancing the classification of pulmonary data sets using generative adversarial networks. Biomedical Research,vol.29,2018,pp3022-3027.
32.Wei Liu, Zhiming Luo, Shaozi Li, Improving deep ensemble vehicle classification by using selected adversarial samples. ELSEVIER,vol.160,2018,pp167-175.
33.Yiting Li, Haisong Huang , Qingsheng Xie, Liguo Yao and Qipeng Chen, Research on a Surface Defect Detection Algorithm Based on MobileNet-SSD,Appl.Sci,vol.8,2018, pp1678-1695
34.JiangyunLi,ZhenfengSu,JiahuiGeng,YixinYin, Real-time Detection of Steel Strip Surface Defects Based on Improved YOLO Detection Network. IFAC,vol.51,2018,pp76–81
35.Xiaohong Sun , Jinan Gu , Rui Huang , Rong Zou,Benjamin Giron Palomares ,Surface Defects Recognition of Wheel Hub Based on Improved Faster R-CNN.Electronics, vol.8,2019,pp481-497.
36.Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi, You Only Look Once:
Unified, Real-Time Object Detection ,CVPR, Las Vegas, NV, USA ,2016.
37.Joseph Redmon, Ali Farhadi,YOLO9000:Better, Faster, Stronger,CVPR,Honolulu,HI, USA,2017.
38.Joseph Redmon, Ali Farhadi ,YOLOv3: An Incremental Improvement,CVPR, Salt Lake City,USA,2018