在光學字元辨識的研究上，單一特徵抽取和單一分類演算法不能達到很低的錯誤率，許多的研究者往結合多種分類架構發展，在這一篇論文中，我們利用結構式特徵與多層分類法架構做光學字元辨識。在抽取結構式特徵方面，從利用細線化處理所獲得的骨架中抽取子筆劃，再從子筆劃中抽出可代表直線和曲線的性質當做特徵。在兩層式分類架構裡，先使用二元樹分類法去降低系統處理時間，然後再利用鬆弛比對法達到較高的辨識率。為了達到多層分類架構的效能，使用訓練資料的方式去或得一門檻值，讓系統在二元分類法階段可以確定字元的可辨識度，是否使用鬆弛比對法辨識。實驗結果顯示所提出的架構有比單一分類法架構有更高的辨識率，所使用的結構式特徵可容忍多種不同的字型。

In the research of optical character recognition, a single feature extraction method and a single classification algorithm cannot yield very low error rate. Many researchers have turned towards the use of complex structures of classification. In this paper, we use structural features are adopted in designing multiclassifier optical character recognition system. Structural feature extraction process, sub-strokes are extracted from the skeleton of a character which is obtained from the thinning process. The strokes extracted from the proposed stroke extraction method include straight and cursive strokes that described with several properties of segment line as the embedding features. In two-layer classification scheme, binary tree classification is used to reduce the processing time and relaxation matching is succeeded to achieve a high recognition. To achieve the optimal performance of multiclassifier, training data is utilize to obtain get the optimum distance threshold value for binary tree classifier. Experimental results show that the performance of the proposed OCR system attains higher recognition rate than the single classifier and the structural features are more tolerant of irregularities in appearing different fonts.

[1]Ivind Due Trier, Anil K. Jain and Torfinn Taxt, “Feature Extraction Methods for Character Recognition - A Survey”, Pattern Recognition, Vol.29, No.4, pp.641-662, 1996.
[2]C. H. Leung and L. Sze, “Feature Selection in the Recognition of Handwritten Chinese Characters”, Engng Applic. Artif. Intell., Vol.10, No.5, pp.495-502, 1997.
[3]Kazuo Toraichi, Tadahiko Kumamoto, Kazuhiko Yamamoto and Hiromitsu Yamada, “Feature analysis of handprinted Chinese characters”, Pattern Recognition Letters, 17(1996)795-800.
[4] 吳偉賢, “筆劃特徵用於離線中文字的辨認”, 國立中央大學資訊工程研究所博士論文, 2000.
[5] L. Heutte, T. Paquet, J. V. Moreau, Y. Lecourtier and C. Olivier, “A Structural / Statistical Feature Based Vector For Handwritten Character Recognition”, Pattern Recognition Letters, 19(1998) 629-641.
[6] Basilios Gatos, Nikos Papamarkos and Christodoulos Chamzas, “Using Curvature Features in a Multiclasifier OCR System”, Engng Applic. Artif. Intell., Vol.10, No.2, pp.213-224, 1997.
[7] Basilios Gatos, Nikos Papamarkos and Christodoulos Chamzas, “A Binary-Tree-Based OCR Technique for Machine-Printed Characters”, Engng Applic. Artif. Intell., Vol.10, No.4, pp.403-412, 1997.
[8] Fang-Hsuan Cheng, “Multi-Stroke Relaxation Matching Method For Handwritten Chinese Character Recognition”, Pattern Recognition, Vol.31, No.4, pp.401-410, 1998.
[9]張鴻楨, “以結構特徵為基礎之鬆弛比對法用於線上手寫中文辨識”, 國立中央大學資訊及電子工程研究所碩士論文, 1993.
[10] Jin Wook Kim, Kwang In Kim, Bong Joon Choi and Hang Joon Kim, “Decomposition of Chinese character into strokes using mathematical morphology”, Pattern Recognition Letters, 20(1999) 285-292.
[11] Ji-Rong Lin and Chang-Fu Chen, “Stroke Extraction For Chinese Characters Using A Trend-followed Transcribing Technique”, Pattern Recognition, Vol.29, No.11, pp.1789-1805, 1996.
[12] 武宜芳, “以正規向量為基礎之手寫中文字筆劃抽取”, 國立中央大學資訊工程研究所碩士論文, 1995.
[13] Chungnan Lee and Bohom Wu, “A Chinese-Character-Stroke-Extraction Algorithm Based On Contour Information,” in Pattern Recognition, Vol. 31. No. 6. Pp. 651-663, 1998
[14] Ben-Kwei Jang and Roland T. Chin, “Analysis of Thinning Algorithms Using Mathematical Morphology”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.12, No.6, June 1990.
[15] Liyuan Li and Weinan Chen, “Corner Detection and Interpretation on Planar Curves Using Fuzzy Reasoning”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.21, No.11, November 1990.