臺灣博碩士論文加值系統

交通運輸建設為國家社會之重要建設之一，其原因在於國家的任何經濟、政治社會等層面的運作都需要透過交通來完成，因此，現代的國家莫不開始戮力研究如何結合電子、資訊、通訊與網路技術來建構「智慧型運輸系統」(Intelligent Transportation System, ITS)。發展ITS的主要目標皆是希望可以改善交通、環保、減少交通擁擠、縮短行車時間，降低肇事率等等。本論文之目的，是要探討如何運用影像處理與電腦視覺等相關技術，來發展一套車牌自動辨識系統，以期為智慧型交通管理建立基礎。

本論文所提出的車牌自動辨識系統，除了在影像前處理上將運用各種影像處理技術來尋找車牌位置與利用樣本比對法( template matching )抽取文字影像外，並將提出以多重專家(Multiple Experts)為主要的辨認架構，以期提高系統的正確率。因此本文將會以快速的方法抽取各種不同的區域、整體統計特徵，並交給兩個專家辨認器進行辨認。專家辨認器是以平行的方式同步辨認，而後再結合兩者之辨認結果得到較準確的辨認答案。

在實驗部分，所使用來辨認的車輛影像，為高速公路一天24小時的車輛影像，所以每一張影像的明暗度都不一樣，因此若是想要使用單一閥值來處理全部影像，是不太可能可以做到的。但是本論文所提系統的車牌定位正確率、字元切割正確率及字元辨認率皆高達97%以上，而整體辨認正確率率則為92.8%以上。

The development of transportation is one of very important strategies in a developing country because the manipulation of economy, politics, and society has to rely it to normally operate. Therefore, modern countries start to investigate how to combine electronics with information, communication and network technologies to build the Intelligent Transportation System (ITS). The purpose of developing ITS is to improve traffic efficiency and environmental protection. The goal of this thesis is to use the automatic technologies of image processing and computer vision to develop an automatic license plate recognition system with an eye to establishing the foundation of intelligent transportation management.

In the proposed license plate recognition system, rule-based technique is first employed to locate license plate. Then, template matching is adopted to extract character image. Last, multiple experts is applied to construct main recognition module for increasing the character recognition rate. In the recognition module, local and global statistic features are extracted by devised methods. These features are then fed to the two experts recognizers. The two experts operate parallelly to recognize the character image. The decisions are combined to yield better recognition result.

Experiments are conducted on which were vehicle images taken from highway day and night. Hence, the brightness of images will not be the same. For this reason, it is impossible to handle all images by a single threshold. However, the license plate locating rate, character segmentation successful rate, and character recognition rate in our system are all over 97%, and the whole recognition rate is 92.8%.

Abstract...........................................................i
摘要.............................................................iii
目錄..............................................................iv
附圖目錄..........................................................vi
第一章 緒論.......................................................1
1.1 研究動機..............................................1
1.2 相關研究..............................................2
1.2.1 文字識別...................................2
1.2.2 車牌辨識...................................5
1.3 系統架構..............................................8
1.4 論文架構..............................................9
第二章 車牌定位..................................................10
2.1 前處理...............................................10
2.1.1 低通濾波..................................11
2.1.2 直方圖伸展................................12
2.1.3 二值化....................................14
2.2 邊緣偵測.............................................15
2.3 車牌定位.............................................17
第三章 文字切割..................................................25
3.1 樣本比對法...........................................25
3.2 文字切割.............................................29
第四章 文字識別與確認............................................34
4.1 文數字訓練及測試樣本取得.............................35
4.2 樣本比對.............................................36
4.2.1 樣本比對專家的學習與訓練..................37
4.2.2 文字識別..................................38
4.3 類神經網路...........................................40
4.3.1 訓練神經網路..............................42
4.3.2 Zernike Moment............................46
4.3.3 文字識別..................................49
4.4 多重專家.............................................51
4.4.1 多重專家結合的類型........................51
4.4.2 多重專家結果的整合........................54
4.5 確認與結果整合.......................................56
第五章 實驗結果..................................................59
5.1 車輛影像樣本的取得...................................59
5.2 車牌辨識系統.........................................61
5.2.1 車牌定位..................................61
5.2.2 文數字切割................................63
5.2.3 文數字辨認................................64
5.3 討論.................................................66
5.3.1 實驗結果..................................66
5.3.2 影像的困難點..............................67
5.3.3 錯誤類型分析..............................67
第六章 結論與未來工作............................................69
6.1 結論.................................................69
6.2 未來工作.............................................71
參考文獻..........................................................72

[1]H. J. Lee, Y. C. Lin, "Using confusing characters to improve character recognition rate", 1998 IEEE International Conference on Systems, Man, and Cybernetics, Vol. 5, 1998,pp. 4195 -4200.[2]R.P.W. Duin, D.M.J. Tax, "Experiments with classifier combining rules", First International Workshop on Multiple Classifier System, Cagliari, Italy, June 2000,Proceedings, p16-29[3]C.Y. Suen, L. Lam, "Multiple classifier combination methodologies for different output levels", First International Workshop on Multiple Classifier System, Cagliari, Italy, June 2000,Proceedings, p52-66[4]L. Lam, "Classifier combinations: Implementations and theoretical issues", First International Workshop on Multiple Classifier System, Cagliari, Italy, June 2000,Proceedings, p77-86[5]A.F.R. Rahman, M.C. Fairhurst, "An evaluation of multi-expert configurations for the recognition of handwritten numerals", Pattern Recognition, Vol.31, No.9, pp.1255-1273,1998[6]A. Khotanzad, Y.H. Hong, (1990), "Invariant image recognition by Zernike moments". IEEE Transactions on Pattern Analysis and Machine Intelligence, 12(5),489-497.[7]G.D Lee, K.S Kim, D.S Jeong, "Rough edge detection of low contrast images using consequential local variance maxima", TENCON 99. Proceedings of the IEEE Region 10 Conference, Volume: 1 , 1999 Page(s): 734 -737 vol.1[8]M. Raus, L. Kreft, "Reading car license plates by the use of artificial neural networks", Circuits and Systems, 1995., Proceedings., Proceedings of the 38th Midwest Symposium on , Volume: 1 , 1996 Page(s): 538 -541 vol.1[9]M.H. ter Brugge, J.H. Stevens, J.A.G Nijhuis, L. Spaanenburg, "License plate recognition using DTCNNs", 1998 Fifth IEEE International Workshop on Cellular Neural Networks and their Applications, London, England, 14-17 April 1998[10]P. Comelli; P. Ferragina; M.N. Granieri; F. Stabile, "Optical recognition of motor vehicle license plates", IEEE Transactions on Vehicular Technology, Volume: 44 Issue: 4 , Nov. 1995 ,Page(s): 790 -799[11]H. A. Hegt, R. J. de la Haye, and N. A. Khan, "A high performance license plate recognition system", IEEE Int. Conf. on Systems, Man, and Cybernetics, pp. 4357-4362, vol. 5, Oct.1998[12]S. H. Park, K. I. Kim, K. Jung, H. J. Kim, "Locating car license plates using neural networks", Electronics Letters Vol. 35 17, 19 Aug. 1999, pp. 1475 -1477[13]J. A. G. Nijhuis, M. H. ter Brugge, K. A. Helmholt, J. P. W. Pluim, L. Spaanenburg, R. S. Venema, and M. A. Westenberg, "Car license plate recognition with neural networks and fuzzy Logic", Proceedings of IEEE International Conference on Neural Networks, pp. 2185-2903, Perth, Western Australia, 1995[14]林泰良, "智慧型車牌定位與字串分割", 國立臺灣大學電機工程學研究所碩士論文, 2000[15]陳麗奾, "在未設限環境下車牌的定位與辨識", 國立臺灣師範大學資訊教育研究所碩士論文,2000[16]溫福助, "類神經網路樣板比對法於車牌字元辨識之研究", 國立臺灣大學電機工程學研究所碩士論文,2000[17]葉怡成，"類神經網路模式應用與實作"，儒林圖書有限公司，2001年4月七版一刷