車牌辨識系統發展了許多年，目前也應用於高速公路、道路與停車場等地方，大多數的車牌辨識系統都是設置於固定的地點，因此也限定了車牌辨識系統的使用範圍，以贓車查緝為例，現在警方是以人工的方式輸入可疑車輛的車牌號碼，藉此判定是否為贓車，假若傳統的車牌辨識系統是可攜帶式的，那便可簡化操作的方式，藉由車牌辨識的方式，自動將辨識結果與贓車資料庫做比對，找出贓車。
為了實現可攜式車牌辨識系統，我們提出了能夠在Android手機系統使用的車牌辨識系統，原始的Android系統並沒有影像處理相關API，我們透過Android將輸入影像解碼，取得raw data，並自行實做所有影像處理演算法，來完成車牌辨識動作。
由實驗的結果可以知道，在Android Dev Phone 2 （ADP2），528Mhz CPU、320萬畫素的照相模組環境下，若使用Pure JAVA方式實做，整個車牌辨識流程大約需要2.45秒，為了進一步提昇執行速度，我們將影像處理相關演算法，透過Native Code的方式實做，並使用JNI技術封裝成library，使用時只要直接呼叫我們的API，就可以完成車牌辨識的動作，最後整個辨識流程只需要0.76秒提昇了3.2倍的速度，也證明了可攜式車牌辨識系統的可行性。

Researches of license plate recognition system have been developed for many years. These systems have been used in highways, roads, parking lots and other places. Most of the license plate recognition systems are installed in fixed locations, therefore it limits the operating range of the system. Take the detection of stolen vehicles as an example, police officers have to manually enter the license plate numbers of suspicious vehicles in order to determine whether the vehicles are the stolen. A portable license plate recognition system will simplify the operations of license plate recognition for this purpose. It can automatically compare the results of license plate recognition system with the stolen vehicle database.
In order to achieve a portable license plate recognition system, we propose a license plate recognition system based on the Google Android mobile phone system. Currently, the Android system does not have APIs for image processing, so we load the picture, decode it and obtain the raw data via the Android System then we use the image processing APIs developed by own lab to complete the license plate recognition.
Our experimental environment is the Android Dev Phone 2 (ADP2) which has 528 MHz ARM11 CPU with 320 megapixel camera module. The license plate recognition process with a pure Java implementation takes about 2.45 seconds. In order to speed up the recognition process, we implement the image processing algorithms with a native implementation and use the JNI technology to pack into image processing library. Using this approach, the entire process requires only 0.76 seconds, and it has a speed-up of about 3.2 compared to the pure Java implementation. So our portable license plate recognition system is feasible to serve the purpose for quick identification of stolen car.

[1] 何坤達，「嵌入式車牌辨識系統」，碩士論文，國立台灣科技大學，台北（2006）。
[2] 李志文，「嵌入式即時多標的汽機車牌照辨識系統」，碩士論文，國立台灣科技大學，台北（2007）。
[3] 黃振欣，「應用倒傳遞網路模型設計PDA即時車牌辨識系統」，碩士論文，大同大學，台北（2004）。
[4] 科技產業資訊室—Kyle。（2009）。市場報導：2009年智慧型手機出貨量成長率6%~11.1%。[Online]。Available: http://cdnet.stpi.org.tw/techroom/market/eetelecomm_mobile/2009/eetelecomm_mobile_09_005.htm
[5] Android Dev Phone. [Online]. Available: http://en.wikipedia.org/wiki/Android_Dev_Phone#Android_Dev_Phone_2
[6] Rob, J. （2009）. Android Dev Phone 2 （ADP2） Is The Google Ion. [Online]. Available: http://phandroid.com/2009/11/13/android-dev-phone-2-adp2-is-the-google-ion/
[7] S. Stobart and M. Vassileiou, PHP and MySQL Manual: Simple, yet Powerful Web Programming, Springer,（2004）
[8] 莊家龍，「車輛偵測與車牌辨識系統」，碩士論文，國立中正大學，嘉義（2005）。
[9] H.J. Lee, S.Y. Chen and S.Z. Wang, “Extraction and recognition of license plates of motorcycles and vehicles on highways,” IEEE Proceedings of the 17th International Conference on Pattern Recognition, vol. 4, pp. 356-359,（2004）
[10] 鐘國亮，影像處理與電腦視覺，第三版，東華書局，台北，（2006）。
[11] S. Hashimi, S. Komatineni and D. MacLean, Pro Android 2, 1st ed., Apress,（2010）
[12] M. D. Fairchild, Color Appearance Models, Addison-Wesley, Reading, MA,（1998）
[13] Color. [Online]. Available: http://msdn.microsoft.com/en-us/library/aa511283.aspx
[14] W. Backhaus, R. Kliegl and J. S. Werner, Color Vision: Perspectives from Different Disciplines, Walter de Gruyter,（1998）
[15] N. Otsu, “A threshold selection method from gray level histogram,” IEEE Trans. On Systems, Man, and Cybernetics, pp. 62-66,（1979）
[16] D. Switkin. ZXing （"Zebra Crossing"）. [Online]. Available: http://code.google.com/p/zxing/
[17] O. Hsueh. WoW ! I see.: Android JNI. [Online]. Available: http://orsonlife.blogspot.com/2009/12/android-jni.html
[18] J.R. Parker, Algorithms for Image Processing and Computer Vision, John Wiley & Sons, Inc., New York, NY,（1996）
[19] E.R. Davies, Machine Vision. Theory, Algorithms, Practicalities, 3rd ed., Morgan Kaufmann, San Francisco,（2005）
[20] R.C. Gonzalez And R.E. Woods, Digital Image Processing, 2nd ed., Prentice-Hall, New Jersey,（2002）
[21] A. Rosenfeld, “Connectivity in Digital Pictures,” Journal of the ACM (JACM), vol. 17 no. 1, pp. 146-160,（1970）
[22] F. Chang, C-J. Chen, and C-J. Lu, “A Linear-Time Component-Labeling Algorithm Using Contour Tracing Technique,” Computer Vision and Image Understanding, vol. 93, no. 2, pp. 206-220,（2004）
[23] M.J. Burton and N.G. Bourbakis, “A comparison of OCRnext term algorithms,” Proc. 9th Int. Conf. on Simulation, vol. 19,（1998）
[24] P. Selinger （2007）. Review of Linux OCR software. [Online]. Available: http://www.mscs.dal.ca/~selinger/ocr-test/
[25] R. Elmasri and S. B. Navathe, Fundamentals of Database Systems, 4th ed., Addison Wesley,（2006）
[26] S. Hashimi, S. Komatineni and D. Maclean, Pro Android 2, 1st ed., Apress,（2010）
[27] M.R. Gupta, N.P. Jacobson and E.K. Garcia, “OCR binarization and image pre-processing for searching historical documents,” Pattern Recognition, vol. 40, no. 1, pp. 389-397,（2007）
[28] K. Wu, E. Otoo and K. Suzuki, “Optimizing two-pass connected-component labeling algorithms,” Pattern Analysis & Applications, vol. 12 no. 2, pp. 117-135,（2009）
[29] C. Fiorio and J. Gustedt, “Two linear time Union-Find strategies for image processing,” Theoretical Computer Science, vol. 154 no. 2, pp. 165-181,（1996）
[30] A. kirillov, AForge.NET Framework. [Online]. Available: http://code.google.com/p/aforge/
[31] bbtesseract. Available: http://code.google.com/p/bbtesseract/
[32] Java Native Interface: Programmer's Guide and Specification. [Online]. Available: http://java.sun.com/docs/books/jni/