基于多层特征融合的名片字符识别方法  

Research on business card character recognition algorithm based on multi⁃layer feature fusion

作　　者：曹仕羽 CAO Shiyu(Department of Infomation Techology,Guodian Nanjing Automation Co.,Ltd.,Nanjing 211100,China)

机构地区：[1]国电南京自动化股份有限公司信息技术事业部,江苏南京211100

出　　处：《电子设计工程》2020年第23期177-182,共6页Electronic Design Engineering

摘　　要：名片识别系统中,字符的字体字号多样性导致归一化后识别率依然较低。在字符识别中,虽然采用Gabor滤波器对字体和光照具有很好的鲁棒性,能准确表征字符的局部纹理特征,但是Gabor特征表征字符图像的全局能力较差。针对上述问题,提出一种多特征融合的字符识别方法,在提取Gabor特征的基础上,融合字符图像的全局特征,使之对字符有更高的区分度。设计了级联分类器进行字符的预测,进一步提高了名片字符识别的速度和精度。In the name card recognition system,the fonts and size of the characters are diverse,resulting in a low recognition rate after character normalization.As for the character recognition,according to Gabor filter is robust to font and illumination changes,it can accurately represent the character’s local texture features.Whereas,for the disability of Gabor filter bank on poor global feature representation.In order to solve the above problems,this paper proposes a character recognition method based on multi feature fusion.On the basis of Gabor feature extraction,the global feature of character image is fused to make it have a higher discrimination for characters.We design a cascaded classifier to predict the characters,which further improves the accuracy and speed of character recognition.

关 键 词：名片识别 字符识别 特征融合 GABOR特征 奇异值特征 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]