臺灣博碩士論文加值系統

在現今的智慧傳輸系統中，車牌辨識是非常重要的一環。車牌辨識廣泛地應用於許多系統上，像是高速公路收費以及停車場管理系統。自2012年起，台灣的交通部公路總局發行了新一代的車牌編碼規則，長度由原本的六碼新增為七碼，且字體也有所更動。為了整合新制車牌與舊制車牌在單一系統上的辨識問題，我們將重新設計一個模型同時兼容兩者，並簡化傳統車牌辨識系統的繁複流程，省去許多影像前處理的動作，來達到相同的效果。
在此篇論文當中，我們基於類神經網路來實作整合六碼及七碼車牌的辨識系統。類神經網路的系統架構及網路訓練使用了Google團隊釋出的開源工具Tensorflow來完成，除了大幅降低以往車牌辨識系統的繁複過程，在辨識準確度及速度都能達到一定水準，此外針對部分遮蔽或是較不清晰的車牌影像，也能夠辨識成功。

License Plate Recognition (LPR) is a very important part of an intelligent transportation system. The LPR is broadly used in many applications, such as highway toll station and car park management. From 2012, Directorate-General of Highway (Taiwan) pursues a new type of license plate, which is changed with the type of font and extra one more digit than existing license plate. In order to solve the recognition on the new license plate (7 digits) and existing license plate (6 digits), we need to redesign a model to made these two types of license plates can be completed in a single system.
In this work, we implement a LPR system based on the neural network to achieve 6 digits and 7 digits’ license plate recognition. The neural network training part is completed with an open source tool, Tensorflow. The results show that our LPR system obtains high accuracy rate even the characters are defaced.

Chapter 1 Introduction 1
1.1. Artificial Neural Network 2
1.2. Organization 3
Chapter 2 Related work 4
2.1. Image Processing 4
2.2. Convolutional Neural Networks 4
2.2.1. Local Connectivity 7
2.2.2. Spatial arrangement 8
2.2.3. Parameter Sharing 9
2.3. Tensorflow of Google tool 10
Chapter 3 System Architecture 12
3.1. 5-Layer CNN Model 13
3.2. Softmax Regression 14
3.3. Implementation of The Regression 17
3.4. Build a Graph 18
3.5. Weights Training 22
3.5.1. The Session 24
3.6. Evaluate Model 25
3.7. Weight initialization 25
3.8. Convolution and Pooling 26
3.8.1. First Convolutional Layer 26
3.8.2. Second Convolutional Layer 27
3.8.3. Densely Connected Layer 27
3.9. Train and Evaluate the model 28
3.9.1. Train Loop 29
3.9.2. Feed the Graph 29
3.9.3. Check the Status 30
3.9.4. Visualize the Status 31
3.9.5. Save Checkpoint 32
3.10. License Plate Detection/Recognition 32
3.10.1. ROI Crop 33
3.10.2. Plate Edges Detection 33
3.10.3. Window Scanning 34
Chapter 4 Experiment and Results 36
4.1. Results 37
Chapter 5 Conclusion 41
5.1. Future work 41
Reference 42

[1]H. R. A. Moghassemi, A. Broumandnia, and A. R. Moghassemi, “Iranian License Plate Recognition using connected component and clustering techniques, in Networked Computing and Advanced Information Management (NCM), 2011 7th International Conference on, 2011, pp. 206–210.
[2]J. Jagannathan, A. Sherajdheen, R. M. V. Deepak, and N. Krishnan, “License plate Character Segmentation using horizontal and vertical projection with dynamic thresholding, in Emerging Trends in Computing, Communication and Nanotechnology (ICE-CCN), 2013 International Conference on, 2013, pp. 700–705.
[3]Y. A. Bolotova, A. A. Druki, and V. G. Spitsyn, “License plate recognition with hierarchical temporal memory model, in Strategic Technology (IFOST), 2014 9th International Forum on, 2014, pp. 136–139.
[4]J.-L. Zhao, H.-G. Min, X.-C. Li, and Y. Pan, “A license plate recognition algorithm under low illumination environment, in 2015 IEEE First International Smart Cities Conference (ISC2), Guadalajara, Mexico, pp. 1–6.
[5]R. Yarisi, E. D. D. Claudio, G. Lucarelli, and G. Orlandi, “Experimental car plate recognition system by neural networks and image processing, in Signal Processing Conference (EUSIPCO 1998), 9th European, 1998, pp. 1–4.
[6]Song Qing-kun, Yuan Hui-jun, and Zhou Teng, “License plate recognition based on mathematical morphology method and RBF neural network, in Proceedings of 2012 International Conference on Measurement, Information and Control, Harbin, China, pp. 782–786.
[7]Y. Cheng, J. Lu, and T. Yahagi, “Car license plate recognition based on the combination of principal components analysis and radial basis function networks, in Signal Processing, 2004. Proceedings. ICSP ’04. 2004 7th International Conference on, 2004, vol. 2, pp. 1455–1458 vol.2.
[8]P. Liu, G. Li, and D. Tu, “Low-quality License Plate Character Recognition Based on CNN, in 2015 8th International Symposium on Computational Intelligence and Design (ISCID), Hangzhou, China, pp. 53–58.
[9]J. Sharma, A. Mishra, K. Saxena, and S. Kumar, “A hybrid technique for License Plate Recognition based on feature selection of wavelet transform and artificial neural network, in Optimization, Reliabilty, and Information Technology (ICROIT), 2014 International Conference on, 2014, pp. 347–352.
[10] I. J. Goodfellow, Y. Bulatov, J. Ibarz, S. Arnoud, and V. Shet, “Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks, arXiv [cs.CV], 20-Dec-2013.
[11] C. Y. Wen, T. S. Huang, C. C. Tseng and S. S. Huang, License plate localization and recognition under different illumination conditions, 2016 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW), Nantou, 2016, pp. 1-2.
[12] TensorFlow [Online]. https://www.tensorflow.org