(3.238.130.97) 您好!臺灣時間:2021/05/18 19:13
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果

詳目顯示:::

: 
twitterline
研究生:郭光瑜
研究生(外文):KUO,KUANG-YU
論文名稱:智慧手機遠端車牌辨識
論文名稱(外文):Remote License Plate Recognition Via Smartphone
指導教授:蔣遐齡蔣遐齡引用關係陳映濃陳映濃引用關係
指導教授(外文):CHIANG,HSIA-LINGCHEN,YING-NONG
口試委員:徐學群金鴻鈞蔣遐齡陳映濃
口試委員(外文):HSU,HSUEH-CHUNCHIN,HUNG-CHUNCHIANG,HSIA-LINGCHEN,YING-NONG
口試日期:2014-05-22
學位類別:碩士
校院名稱:醒吾科技大學
系所名稱:資訊科技系所
學門:電算機學門
學類:電算機一般學類
論文種類:學術論文
論文出版年:2014
畢業學年度:102
語文別:中文
論文頁數:119
中文關鍵詞:智慧型手機車牌辨識影像定位字元辨識AdaBoost演算法
外文關鍵詞:Smart PhoneLicense Plate RecognitionImage ProcessingApplicationsAdaBoost Algorithm
相關次數:
  • 被引用被引用:2
  • 點閱點閱:614
  • 評分評分:
  • 下載下載:164
  • 收藏至我的研究室書目清單書目收藏:0
台北市政府於2011年1月5日正式啟用其花費13億4800萬建置13699支固定監視器的全市遠端錄影監控系統(Telematics System);此系統內建車牌自動辨識技術,可用以強化政府交通管理與改善治安[40]。但僅靠政府於定點設立有限的固定監視器,其功能似乎不足,應該使用行動裝置提供全民參與進行補強。根據Google委託台灣易普索(Ipsos Taiwan)公司所作調查,台灣地區2013年智慧手機普及率已高達51% [41],將此資源有效運用在全台灣超過2200萬台[42]機動車輛之監控與查察,應屬可行。
智慧型手機應用於車牌辨識,除了車牌字元之辨識成功率外,其處理速度、使用便捷性及原始影像保全等亦為實用化之考量重點。本研究使用分散式處理設計,將智慧手機所攝取的車牌原始影像先進行影像定位處理,並降低其輸出資料量後,再透過3.5G無線通信傳輸至後端主機進行字元辨識處理及車籍資料庫比對,藉以發掘整體系統之技術瓶頸所在,建構出可行的系統架構。
本研究之影像定位及字元辨識處理皆使用AdaBoost演算法。智慧手機執行的影像定位處理是要產生只有牌照區域的最小灰階影像,以利3.5G無線通信快速傳輸,後送至雲端伺服器進行平行處理及車籍資料比對,滿足瞬間鉅量及高速運算之需求。系統開發之程序是先在個人電腦運用OpenCV模式識別程序之AdaBoost分類器進行牌照區域辨識訓練,以建構完整之車牌辨識系統並測試其最小可辨識之牌照資料量,再將牌照定位程式與訓練成果及最小可辨識之牌照資料量編寫成JAVA程序寫入智慧型手機、字元辨識與訓練成果寫入系統主機,進行測試。實務驗證是使用Android智慧手機作為前端攝像及影像定位及處理平台,透過應用程式(APP)之開發進行全自動化作業;後端字元辨識平台則使用獨立之電腦主機與智慧手機進行通聯。
實驗顯示,本系統架構在沒有影像正向化處理及車籍資料庫比對校正之狀況下,單張原始影像從取樣至辨識完成所費時間少於0.1秒,辨識成功率為88.8%,初步可滿足實用化之需求;但智慧手機之攝像環境遠較固定監視器複雜,致使其所攝影像之定位處理不易,產生實務操作上的困難、定位處理費時及辨識成功率下降等問題。改善的方法是使用自動化動態連拍攝影處理及影像正向化處理技術,但這也將造成智慧手機運算量大幅增加,使得影像定位及字元辨識分散處理之系統架構成為必需。
關鍵字:智慧型手機、車牌辨識、影像定位、字元辨識、APP、AdaBoost演算法

On January 5, 2011, Taipei Municipal Government officially initiated the city-wide remote video surveillance system comprising 13,699 fixed monitors built at a cost of NTD 1.3 billion and 48 million. The surveillance system includes an automatic license plate recognition technology and can be used to strengthen the governmental traffic management and improve public security. However, the function of the limited fixed monitors provided at specific places by the government seems insufficient and should be enhanced by the use of mobile devices to provide universal participation. According to the survey of Ipsos in Taiwan commissioned by Google, the popularization rate of smart phones in Taiwan area in 2013 reached 51%. Therefore, it should be feasible to use this resource effectively throughout Taiwan for monitoring and inspecting more than 22 million units of motor vehicles.
In addition to the success rate of character recognition of license plates, when smart phones are used for license plate recognition, their processing speed, ease of use and preservation of original images are priorities to be considered for practical application. In this study, a distributed processing design is adopted. In this process, first, the original images of license plates taken by smart phones are processed in image positioning and then, with the amount of output data reduced, they are transmitted via 3.5G wireless communications to the back-end mainframe for character recognition and vehicle registration database comparison in order to explore the technical bottleneck of the whole system and construct a workable system configuration.
AdaBoost algorithm is used for both the image positioning and character recognition and processing in this study. The purpose of smart phone image positioning processing is to produce the smallest grayscale images of only license areas to facilitate rapid transmission via 3.5G wireless communications, and then transmit them to the cloud server for parallel processing and vehicle registration data comparison, as well as to meet the demand of massive and instantaneous high speed computation. The procedure for the system development is to use the AdaBoost classifier of OpenCV pattern recognition program on personal computers to have a training on license plate area recognition to construct a complete license plate recognition system and test the amount of its minimum recognizable license data, and then compile the license positioning program, the training results, and the amount of its minimum recognizable license data into JAVA program to be written into smart phones, while character recognition and training results are written into the system mainframe for testing. The practical validation is conducted using Android smart phones as the front-end video recording, image positioning, and processing platform for fully automated operations through the development of applications, while the back-end character recognition platform will use an independent computer mainframe to communicate with smart phones.
Experiments show that, in the system configuration, under the conditions of no positive image processing and vehicle registration database comparison and correction, the time from sampling to recognition completion of a single original image is less than 0.1 seconds, and the rate of successful recognition is 88.8%, which can meet the initial practical needs. However, the video recording environment of smart phones is far more complex than fixed camera monitors, resulting in a difficult positioning processing of the images taken, thus producing difficulties in practical operation such as time-consuming positioning processing, reduction in rate of successful recognition, and other issues. Using an automated processing of dynamic continuous shooting and positive image processing technology can help improve this. However, it will also result in a substantial increase in the amount of smart phone operations, making it necessary to have the system configuration for image positioning and distributed processing character recognition.

Keywords: Smart Phone, License Plate Recognition, Image Processing, Applications, AdaBoost Algorithm

摘 要 i
ABSTRACT ii
誌 謝 iv
目 錄 v
表 目 錄 vii
圖 目 錄 viii
1. 緒論 1
1.1. 研究動機 1
1.2. 研究目的 3
1.3. 研究流程 4
1.4. 研究範圍 6
1.5. 研究問題探討 8
2. 文獻回顧 11
2.1. 車牌定位 11
2.1.1 邊緣偵測 11
2.1.2 模糊理論 12
2.1.3 形態學 12
2.1.4 AdaBoost演算法 13
2.2. 字元分割 14
2.2.1. 連通元件法 14
2.2.2. 投影法 15
2.3. 字元辨識 16
2.3.1. 結構法 16
2.3.2. 樣板比對法 16
2.3.3. 類神經網路法 17
2.3.4. AdaBoost演算法 18
2.4. 雲端簡介 19
2.4.1. 雲端運算服務 19
2.4.2. Windows Azure 23
2.4.3. 雲端開發 24
2.5. 影像前置處理 27
2.5.1. 影像灰階化 28
2.5.2. 影像二值化 30
2.5.3. 平滑濾波器 30
2.5.4. 影像正向化 30
3. 研究架構設計 31
3.1. 核心問題與研究架構 31
3.1.1. 智慧手機前端影像處理 31
3.1.2. AdaBoost演算法 32
3.1.3. 研究架構 34
3.2. 研究設計 35
3.2.1. 影像灰階化 39
3.2.2. AdaBoost演算法 42
3.3. 研究方法 43
3.3.1. 研究方法 43
3.3.2. 研究工具 43
4. 驗證方法設計 51
4.1 最小可辨識牌照影像 51
4.2 智慧手機結合混合雲之車牌辨識系統 54
5. 驗證測試 57
5.1. 環境建置 57
5.2. 軟硬體介紹 58
5.2.1. 硬體 58
5.2.2. 軟體 60
5.3. 測試方式 62
6. 分析與討論 65
6.1 數據分析 65
6.2 研究討論 68
7. 結論與建議 83
8. 參考文獻 85
附錄A 實驗樣本(■成功;■失敗) A-1

[1]R.C. Gonazlez, R.E. Woods, S.L. Eddins (2002),Digital Image Processing, using MATLAB, Prentice Hall.
[2]M. Sonka, V. Hilavac, R. Boyle (1998), Image Processing, Analysis, and Machine Visionm, PWS Publishing.
[3]Zeidenberg, Matthew (1990), Neural Networks in Artificial Intelligence.Ellis Horwood Limited.
[4]Y. Freund and R. E. Schapire, A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting, Journal of Computer and System Sci-ences, vol. 55, pp. 119-139, 1997.
[5]R.K.Chellappa & A.Gupta(2002), Managing computing resources in active intranets, International Journal Of Network Management 12(2), pp.117-128.
[6]Louka Dlagnekov(2004), License Plate Detection Using AdaBoost, Department for Computer Science & Engineering, La Jolla.
[7]Capar, A. and M. Gokmen (2006), Concurrent segmentation and recognition with shape-driven fast marching methods,in Proc. 18th Int. Conf. on Pattern Recognition, Hong Kong, Aug.20-24, vol.1,pp.155-158.
[8]Wang, S.-Z., and H.-J. Lee (2003), Detection and recognition of license plate characters with different appearances, in Proc. Conf. Int. Trans. Systems, 2003.,vol.2, pp.979-984, Oct.
[9]Zheng, D., Y. Zhao, and J. Wang (2005), An efficient method of license plate location, Pattern Recognition Letters, vol.26, no.15,pp.2431-2438, Nov.
[10]Bai, H. and C. Liu (2004), A hybrid license plate extraction method based on edge statistics and morphology, in Proc. 17th Int. Conf. on Pattern Recognition, Hangzhou, China, Aug.23-26, pp.831-834.
[11]Mei, Y. and Y. D. Kim (2000), An approach to Korean license plate recognition based on vertical edge matching,in Proc. Int. Conf. on Systems, Man, and Cybernetics, Nashville, TN, Oct.8-11, pp.2975-2980.
[11]S.-L. Chang, L.-S. Chen, Y.-C. Chung and S.-W. Chen (2004), Automatic license plate recognition,IEEE Trans. on Intelligent Transportation Systems, vol. 5, pp.42-53,Mar.
[12]Freund, Y., & Schapire, R. E. (1995). A decision-theoretic generalization of on-line learning and an application to boosting. Second European Conference on Computational Learning Theory, 23–37.
[13] H. Al -Yousefi, S. S . Udpa (1992), Recognition of Arabic Characters, IEEE Trans. Pattern Analysis and Machine Intelligence, 14, 853-857.
[14]M.M.M. Fahmy (1994),Automatic Number-plate Recognition : Neural Network Approach,Proceedings of VNIS’94 Vehicle Navigation and Information System Conference, 3 1 Aug-2 Sept.
[15]Rahman, C. A., Badawy, W., and Radmanesh, A. (2003), A Real Time Vehicle’s License Plate Recognition System,Proceedings of IEEE Conference on Advanced Video and Signal Based Surveillance, pp. 136-166.
[16]Hegt, H. A., de la Haye, R. J., and Khan, N. A. (1998), A High Performance License Plate Recognition System,Proceedings of IEEE International Conference on Systems, Man and Cybernetics, Vol. 5, pp. 4357-4362.
[17]Sebastian Zander,Grenville Armitage,Thuy Nguyen,Lutz Mark,Brandon Tyo(2006), Minimally Intrusive Round Trip Time Measurements Using Synthetic Packet-Pairs, Centre for Advanced Internet Architectures (CAIA). Technical Report 060707A Swinburne University of Technology Melbourne, Australia.
[18]P. Viola and M. Jones(2001), Robust Real-time Object Detection, ICCV 2nd International Workshop on Statistical and Computational Theories of Vision : Modeling, Learning, Computing and Sampling, July.
[19]謝邦昌、邱志洲(2000),類神經網路分析(Neural Network Analysis),曉園出版社。
[20]Wuzhimin(2010),Android之父Andy Rubin:生而Geek,程式員雜誌,第9期。
[21]林嶺、張力為 (2007),運動性心理疲勞問題的研究現狀,心理科學進展期刊,15(3),524~ 531頁。
[22]C. K. Chen (2005), Segmenting License Plate in Color Images, Thesis for Department of Mechanical & Electro-Mechanical Engineering of National Sun Yat Sen University.
[23]周俊男(1995), 車輛牌照辨識影像處理系統,國立中山大學資訊工程研究所碩士論文,高雄市。
[24]張銘豪(1996),利用分割辨識方法之英文數字辨識系統,國立中山大學資訊工程研究所碩士論文,高雄市。
[25]杜宛玲(2008),可以支援公共安全應用之通用型隨插即用技術-以車牌辨識為例,樹德科技大學研究所碩士論文,高雄市。
[26]翁偉鐘(2008),利用Adaboost演算法於數位影像中進行文字定位,亞洲大學,台中市。
[27]王振興 (2003),多標的汽機車車牌辨識系統之研究,元智大學碩士論文,桃園縣。
[28]魏銪志 (2000),動態多標的車牌辨識系統之研究,元智大學碩士論文,桃園縣。
[29]黃琇晴 (2007),應用於道路監視器之自動車牌辨識系統,朝陽科技大學碩士論文,台中市。
[30]李國彥 (2004),車牌辨識系統之研究,大同大學碩士論文,台北市。
[31]張簡子介 (2002),用小腦模型在FPGA 上作車牌辨識,國立台灣師範大學碩士論文,台北市。
[32]呂炎洲 (2004),不需字元切割的車牌辨識法,靜宜大學碩士論文,台中市。
[33]溫福助 (2000),類神經網路樣板比對法於車牌字元辨識之研究,台灣大學碩士論文,台北市。
[34]曾國恭 (2009),以雲端運算服務建構之數位學習管理系統2.0,中原大學碩士論文,桃園縣。
[35]曹培彥 (2005),利用灰階影像直方圖偵測人眼,義守大學碩士論文,高雄市。
[36]林欣平 (1999),車牌字粹取,國立交通大學碩士論文,新竹市。

[37]謝昆霖、鄭秀慧 (2006),應用類神經網路建構成本效益決策支援模式,中華管理評論國際學報,9(4),5頁。
[38]林立 (2010),以雙核心處理器為基礎之車牌辨識系統,國立中央大學碩士論文,桃園縣。
[39]孫中山:三民主義,1905年11月26日,民報。
[40]電眼無敵 北市搶案100%偵破,中時電子報,2013年4月27日,http://news.chinatimes.com/society/110503/112013042700128.html。
[41]Google:台灣智慧型手機普及率已達51%,iThome Online,2013年8月13日,http://www.ithome.com.tw/itadm/article.php?c=82029。
[42]機動車輛登記數,交通部,2013年4月10日,http://stat.motc.gov.tw/mocdb/stmain.jsp?sys=100。
[43]Eucalyptus,中央研究院資訊科學研究所,2014年4月1日,http://www.openfoundry.org/tw/resourcecatalog/Cloud-Computing/Infrastructure-Layer/Computing/eucalyptus。

QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top