一种快速名片字符识别算法  

Fast algorithm of namecard character recognition

作　　者：刚亚州 黄元元[1] 戴群[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,南京210016

出　　处：《计算机应用研究》2014年第9期2859-2863,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61100108)

摘　　要：针对嵌入式名片识别系统中低质名片字符特征提取困难导致识别速度慢和识别率低的问题,提出使用字符加权模板和基于统计的字符特征提取算法相结合,对提取的特征进行编码匹配。该算法对字符模板进行加权操作和网格化处理,并对每一块网格区域提取特征,并将提取出的特征在特征空间进行多尺度划分并采用二进制编码,这样用一串编码表示特征,最后通过编码匹配实现特征匹配。实验结果表明,该模板加权和特征提取方法与特征编码匹配结合可以较大地提高名片字符识别率。To solve the problem of the low character recognition rate and slow speed resulted from the poor quality of the char- acter image in the name card recognition system, this paper proposed a character recognition algorithm of combining feature weighting template and character feature extraction algorithm based on statistics and coding feature to match. The algorithm firstly weighted the character template and divided segmented into many blocks. Secondly, this paper extracted feature of each block grid area and divided into several sub-blocks under different resolutions. And then it assigned each sub-block to an only code. So that those feature points could be represented by the codes. Thirdly, the feature matching could be completed through code matching. The experimental results show that the combining template weighted ,feature extraction and coding matching re- placed feature matchin~ method can be ~reatlv imoroved character recognition rate.

关 键 词：名片识别 特征提取 空间尺度划分 二进制编码 特征匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP301.6[自动化与计算机技术—计算机科学与技术]