本研究主要探討傾斜車牌對於車牌辨識之準確度影響，在傾斜之車牌進行字元分割，容易造成分割不完整或超出範圍，而導致字元辨識準確率下降。有鑑於此，本研究提出一種車牌校正之方法，有效提高車牌辨識準確率且減少訓練時間。本研究方法分為四個步驟：車牌定位、車牌校正、字元分割和字元辨識。車牌定位使用YOLOv4定位車牌之位置並放大，減少了雜訊與背景。車牌校正先使用水平投影，分析每個角度之像素變化，進一步判斷車牌水平傾斜角度並校正。再使用垂直投影，使車牌之上半部像素與下半部像素往反方向移動，分析每個移動角度之投影狀況，進一步判斷車牌垂直傾斜角度並校正。第三步，將校正後之車牌進行字元分割，將每一個字元提取出來，進行字元辨識。第四部字元辨識，使用CNN卷積神經網路，學習超過9000張字元圖片，並經過最大池化層之特徵篩選，訓練預測模型。最後進行比較，未校正之車牌之辨識準確率在25次訓練週期後為96.1％，而校正後之車牌辨識準確率僅在10次訓練週期後為99％，在提高準確率之條件下也縮短了訓練時間，對於傾斜之車牌辨識率有效提高。本研究在進行車牌傾斜校正後與校正前，車牌辨識準確率從96.1％提高為99％，訓練時間從25個訓練週期減少至10個訓練週期，大幅將低了訓練時間。

The purpose of the study was to discuss how tilted license plates affects the accuracy of license plate recognition and how to improve such recognition system. The character segmentation on tilted license plates usually causes character segmentation incomplete or out of the range, which leads to a decrease in the accuracy rate of the character recognition. In this paper proposes a method that improves the accuracy of license plate recognition and reduces the prediction model training time for the recognition system. The study has four steps which are license plate location, license plate correction, character segmentation, and character recognition.
Firstly, license plates were located and zoomed in with YOLOv4 to reduce irrelevant noise and background value. Secondly, the system analyzed pixel changes of each angle with horizontal projection and corrected the horizontal tilt angle for the license plate. Then the system used vertical projection to move the upper and lower half pixels of the license plate in opposite directions. By analyzing the projection status of each angle, the system then corrected the vertical tilt angle for the license plate. Thirdly, the system performed character segmentation on the corrected license plate. This is being done by extracting each character. Lastly, by given more than 9,000 characters pictures from step three, the recognition system with Convolutional Neural Network(CNN) trained the prediction model with the feature selection of the maximum pooling layer.
Finally, the recognition system accuracy of predicting the uncorrected license plate is 96.1% after 25 epochs, while the recognition accuracy of predicting corrected license plate is 99% after 10 epochs. The accuracy of license plate recognition was increased from 96.1% to 99% after license plate tilt correction. CNN training time was decreased from 25 epochs to 10 epochs.

中文摘要 ........................ i
Abstract........................ ii
誌謝........................ iii
目錄........................ iv
表目錄........................ v
圖目錄........................ vi
符號說明........................ viii
第一章 緒論........................ 1
1.1 前言........................ 1
1.2 研究動機與目的........................ 1
1.3 國內外論文回顧........................ 1
1.4 論文架構........................ 2
第二章 車牌辨識與校正模型原理........................ 3
2.1 YOLOv4........................ 3
2.2 水平與垂直投影........................ 4
2.3 卷積神經網路........................ 5
第三章 實驗數據收集與車牌辨識模型建立方法........................ 7
3.1 數據收集與標準環境........................ 7
3.2 車牌定位........................ 8
3.3 車牌校正........................ 9
3.3.1 水平校正........................ 9
3.3.2 垂直校正........................ 11
3.4 字元分割........................ 13
3.5 字元辨識........................ 14
3.6 實驗設備........................ 14
第四章 實驗結果呈現........................ 15
4.1 車牌定位........................ 15
4.2 車牌校正........................ 16
4.2.1 水平校正........................ 16
4.2.2 垂直校正........................ 18
4.3 字元分割........................ 20
4.4 字元辨識........................ 22
第五章 結論與討論........................ 24
參考文獻 ........................ 25
Extended Abstract........................ 26

[1] Chang, Shyang-Lih, et al. "Automatic license plate recognition." IEEE transactions on intelligent transportation systems 5.1 (2004): 42-53.
[2] Ren, Shaoqing, et al. "Faster r-cnn: Towards real-time object detection with region proposal networks." Advances in neural information processing systems 28 (2015).
[3] Zou, Li, et al. "License plate detection with shallow and deep CNNs in complex environments." Complexity 2018 (2018).
[4] Laroca, Rayson, et al. "An efficient and layout‐independent automatic license plate recognition system based on the YOLO detector." IET Intelligent Transport Systems 15.4 (2021): 483-503.
[5] Hsu, Gee-Sern, et al. "Robust license plate detection in the wild." 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS). IEEE, 2017.
[6] Selmi, Zied, Mohamed Ben Halima, and Adel M. Alimi. "Deep learning system for automatic license plate detection and recognition." 2017 14th IAPR international conference on document analysis and recognition (ICDAR). Vol. 1. IEEE, 2017.
[7] Li, Hui, Peng Wang, and Chunhua Shen. "Toward end-to-end car license plate detection and recognition with deep neural networks." IEEE Transactions on Intelligent Transportation Systems 20.3 (2018): 1126-1136.
[8] Kessentini, Yousri, et al. "A two-stage deep neural network for multi-norm license plate detection and recognition." Expert systems with applications 136 (2019): 159-170.
[9] Saif, Nazmus, et al. "Automatic license plate recognition system for bangla LP using convolutional neural network." TENCON 2019-2019 IEEE Region 10 Conference (TENCON). IEEE, 2019.
[10] Pustokhina, Irina Valeryevna, et al. "Automatic vehicle license plate recognition using optimal K-means with convolutional neural network for intelligent transportation systems." Ieee Access 8 (2020): 92907-92917.
[11] Pan, Mei-Sen, Jun-Biao Yan, and Zheng-Hong Xiao. "Vehicle license plate character segmentation." International Journal of Automation and Computing 5.4 (2008): 425-432.
[12] Bochkovskiy, Alexey, Chien-Yao Wang, and Hong-Yuan Mark Liao. "Yolov4: Optimal speed and accuracy of object detection." arXiv preprint arXiv:2004.10934 (2020).
[13] Hubel, David Hunter, and Torsten Nils Wiesel. "Ferrier lecture-Functional architecture of macaque monkey visual cortex." Proceedings of the Royal Society of London. Series B. Biological Sciences 198.1130 (1977): 1-59.
[14]（日）谷岡廣樹．康鑫著，2019，深度學習入門教室，莊永裕譯，臉譜（城邦）文化出版社，台北。