摘要
现有基于字符形状的识别算法依赖其他辅助方法识别字符,对此提出改进。采用背景赋值描述字符的凹凸性特征,提出凹凸对称性和凹凸区域夹角的计算方法作为补充,充分利用字符的凹凸性特征设计多级分类器,完全采用凹凸性特征进行多级分类识别,正确识别25个大写英文字符和10个阿拉伯数字,不受字符大小与倾斜影响。实验结果表明,在此提出的改进算法克服了现有算法缺陷,扩展了凹凸性特征的可区分范围,突出了基于形状的算法的优越性。
An improvement is proposed to avoid relying on assistant methods on which the existing recognition algorithm based on character shape relies.The new algorithm adopts background assignment to describe the convex-concave feature.As supplementary,a calculation method of the convex-concave symmetry and convex-concave included angle is proposed to constitute multi-stage classifier by making full use of convex-concave feature.It employs the convex-concave feature to conduct multistage classification recognition,by which 25 capital English characters and 10 Arabic numerals can be identified correctly.The recognition is not affected by size and tilt of character.The experimental results show that,the new algorithm can expand the distinguish range of convex-concave feature and highlight superiority of the algorithm based on character shape by overcoming the defects of existing algorithm.
出处
《现代电子技术》
2013年第16期77-80,共4页
Modern Electronics Technique
基金
内蒙古工业大学科研基金(200602)
关键词
字符识别
形状特征
凹凸性特征
凹凸对称性
凹凸区域夹角
character recognition
shape feature
convex-concave feature
convex-concave symmetry
included angle of convex-concave
