一种有效的多字体印刷字符识别系统实现  

Implementation of effective multifont printed character recognition system

作　　者：赵全友[1] 潘保昌[1] 姚锦秀[1] 郑胜林[1] 陈箫枫[1] 

机构地区：[1]广东工业大学信息工程学院,广东广州510640

出　　处：《重庆邮电学院学报（自然科学版）》2005年第4期451-455,共5页Journal of Chongqing University of Posts and Telecommunications(Natural Sciences Edition)

摘　　要：提出了一种基于神经网络和浮动模板的多字体印刷字符识别方法。在研究大量的多字体印刷字符图像后,给出了一种有效的预处理方法,并在综合抽取宏观特征与微观特征后,送入神经网络的浮动模板法分类器进行识别。实验证明该方法具有相当高的识别率,应用前景十分广泛。A multifont printed character recognition based on floating mask method and neural network is presented in this paper. After researching on many multifont printed character images, authors give an effective preprocessing method. Then the macro features and the micro features are extracted, which are to be sent to the BP neural network. The classifier of combination of the neural network and the floating mask yields the right results of classification.The experiment proves this method can get very high recognition rate.The prospect of its application in many fields is obvious.

关 键 词：多字体识别 预处理 特征提取 神经网络:浮动模板法 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]