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.
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)
关键词
多字体识别
预处理
特征提取
神经网络:浮动模板法
multifont character recognition
preprocessing
feature extracting
neural network
floating mask method
