光學字元辨識 OCR (Optical Character Recognition)系統應用相當廣泛，例如車牌辨識。傳統的作法是針對每一個字元的特性去做資料庫，辨識時使用資料庫比對，找出差異值最少的，因此延伸出許多產生字元特性的演算法。本論文主要在研究使用投影(Projection)與樹狀決策(Tree Division)在光學字元辨識系統，並以SOPC(System on Programmable Chip)實現。
本文辨識字元的方法，是以投影加上樹狀決策為主軸；辨識字元首要工作為字元切割，將一串字中找出字元個體，再利用字體的投影分布，加上樹狀決策辨別出字元。利用字元每一行與列的投影量，觀察投影分布的情形給予分類，採用樹狀決策分辨出字元，不需建立資料庫，節省大量記憶體，嵌入式系統中記憶體是相當珍貴。利用Harr濾波器做小波(Wavelet)轉換將大的字元縮在自訂的處理範圍，不但節省正規化所需付出的複雜度，且方便硬體實作。
SOPC( System On Programmable Chip) 為一個FPGA，內含有一個32-bit RISC處理器，可以驗證演算法並可加入硬體模組驗證演算法硬體架構
關鍵字：OCR , SOPC , Projection , Tree-decision , wavelet

Optical Character Recognition(OCR) system has widespread applications. Traditional method requires database for special characteristic of each character. The feature of character matches with feature of database. So, The wide range of shape variations for OCR requires an adequate representation of the discriminating features for classification. Proposed research uses projection and tree-decision for character recognition and is implement by Altera’s SOPC(System On Programmable Chip).

Proposed method uses projection and tree decision mainly. First stage of OCR is character segmentation. It finds out individual character from sequence of character and recognizes character by projective distribution and tree decision. The projective distribution is used to classified by tree-decision. It has no database and memory. Memory is valuable in Embedded system. Proposed method uses DWT that the large size character can be reduced into my process range and save complexity of normalization. Implement hardware is easy.
Altera’s SOPC is FPGA and can synthesize 32-bit RISC processor . It is convenient to verify algorithms and hardware architecture.

Index Term- OCR, SOPC, Projection, Tree-decision, wavelet

[1]Wang, P. S. P., World Scientific, “Character & handwriting recognition : expanding frontiers,” World Scientific Series in Computer Science-Vol. 30,1991
[2]Mori S., Suen C.Y., Yamamoto K., “Historical review of OCR research and development,” Proceedings of the IEEE Volume 80, Issue 7, July 1992 pp. 1029 - 1058
[3]Casey R.G., Lecolinet E., “A survey of methods and strategies in character segmentation,” Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 18, Issue 7, July 1996 pp. 690 - 706
[4]Chang T., Kuo C.-C.J., “Texture segmentation with tree-structured wavelet transform,” Time-Frequency and Time-Scale Analysis, Oct. 1992 pp. 543 – 546
[5]Chang T., Kuo C.-C.J, “Texture analysis and classification with tree-structured wavelet transform,” Image Processing, Oct. 1993 pp.429 - 441
[6]Yuan Y. Tang, Bing F. Li, Hong Ma, Jiming Liu, Leung C.H., Ching Y. Suen, “A novel approach to optical character recognition based on ring-projection-wavelet-fractal signatures, ”Pattern Recognition, 25-29 Aug. 1996 pp. 325 - 329 vol.2
[7]DeBrunner V., Kadiyala M., ”Effect of wavelet bases in texture classification using a tree-structured wavelet transform,” Signals, Systems, and Computers, 24-27 Oct. 1999 pp. 1292 - 1296 vol.2
[8]Yang Jun, Yu Songyu, Zhao Rongchun, “Wavelet analysis for handwritten Chinese character recognition,” Intelligent Processing Systems, 28-31 Oct. 1997 pp. 1023 - 1026 vol.2
[9]Gllavata J., Ewerth R., Freisleben B., “Text detection in images based on unsupervised classification of high-frequency wavelet coefficients,” Pattern Recognition, 23-26 Aug. 2004 pp. 425 - 428 Vol.1
[10]Ha T.M., ”Efficient detection of abnormalities in large OCR databases, ” Document Analysis and Recognition, 18-20 Aug. 1997 pp. 1006 - 1010 vol.2
[11]Tellache M., Sid-Ahmed M., Abaza B., “Thinning algorithms for Arabic OCR,” Communications, Computers and Signal Processing, IEEE Pacific Rim Conference on Volume 1, 19-21 May 1993 pp. 248 – 251
[12]Jang B.-K., Chin R.T., ”Analysis of thinning algorithms using mathematical morphology,” Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 12, Issue 6, June 1990 pp. 541 – 551
[13]Gllavata J., Ewerth R., Freisleben B., “A text detection, localization and segmentation system for OCR in images,” Multimedia Software Engineering, IEEE Sixth International Symposium on 13-15 Dec. 2004 pp. 310 – 317
[14]Sawaki M., Hagita N., “Text-line extraction and character recognition of document headlines with graphical designs using complementary similarity measure,” Pattern Analysis and Machine Intelligence, IEEE Transactions on Volume 20, Issue 10, Oct. 1998 pp. 1103 – 1109
[15]Tsujimoto S., Asada H., “Major components of a complete text reading system,” Proceedings of the IEEE Volume 80, Issue 7, July 1992 pp. 1133 - 1149
[16]張仁豪， “正交軸投影法與樹狀決策在汽車牌照辨識的研究，”， 國立清華大學，原子能原學系碩士論文 ，民國90年，六月
[17]Michael D.Ciletti, Advanced Digital Design with the Verilog HDL, Prentice Hall,May2001