離線手寫數字辨識系統的有效區域特徵擷取技術之研究

A Study of Effective Region-features Extraction for Off-line Handwritten Numeral Recognition

李聖偉(Shen-Wei Lee)

指導教授 ： 吳憲珠

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2012年)

摘要

在本論文中，主要探討離線手寫數字辨識之研究。離線手寫數字辨識已被研究多年，雖然已有高辨識率，但仍有訓練之問題。本研究重點能改善上述缺點，提高辨識率。本研究採用國小學生之手寫字樣本做為測試資料，特徵擷取採用SVM，取得數字之區域特徵，以提高辨識率。
首先，本論文在離線離線手寫數字辨識系統方面，利用數位影像處理，包括影像二值化，進行影像前置處理。再經過圖形細線化，計算其總端點數、交叉點數、線條數之技術運用於影像切割三區域之處理，再除去交叉點後，取影像之線條數，方向鏈碼數，做為辨識用之特徵值；依一般型態則判斷字元圖形之各區之方向鏈碼標準差、方向鏈碼之平均、上中下各點之比例、字元圖形之各項點數分布，加上字元圖形細線化後之端點數等結構式特徵，最後再使用SVM之方法，取得更有效的多特徵值。

關鍵字

影像辨識 ； 影像前置處理 ； 結構式特徵

並列摘要

A multiple features extraction technique for the recognition of handwritten numbers is proposed. The proposed technique mainly extracts direction information from the structure of contours of each handwritten number and the direction information is integrated with a technique for detecting transitions among pixels and counting the number of cross lines in the lined image of offline handwritten numbers. The technique used in the recognition combining with a Support Vector Machine (SVM) classifier provides recognition rates up to 98.99%. This proposed technique also uses SVM for determining the effective features extracted from the multiple features of the handwritten number recognition.

並列關鍵字

image recognition ； structural feature ； feature extraction

參考文獻

[1] B. Verma, M. Blumenstein and M. Ghosh, “A novel approach for structural feature extraction: contour vs. direction,” Pattern Recognition Letters, Volume 25, Issue 9, 2004, pp. 975–988.

[2] J. Ni, C. Zhang and S. X. Yang, “An adaptive approach based on KPCA and SVM for real-time fault diagnosis of HVCBs,” IEEE Transactions on Power Delivery, 2011, pp. 1960–1971.

[3] D. Y. C. Leung and C. H. Liu, “Limitations of the relative standard deviation of win percentages for measuring competitive balance in sports leagues, ” Economics Letters, Volume 109, Issue 1, October 2010, pp. 38-41.

[4] J. Melendez, M. A. Garcia, D. Puig and M. Petrou, “Unsupervised texture-based image segmentation through pattern discovery,” Computer Vision and Image Understanding, Volume 115, Issue 8, 2011, pp. 1121-1133.

[5] Y. Wena and L. He, “Classifier for Bangla handwritten numeral recognition,” Expert Systems with Applications, Volume 39, 2012, pp. 948–953.