摘要
针对表格中常用的印刷体汉字提出了一种识别方法 ,并设计与实现了基于该方法的印刷体汉字识别系统 .该系统首先对输入的汉字图像预处理 ,得到规格化的汉字图像 ;然后利用汉字图像的面积进行第 1级中心提取粗分类 ;再利用人工神经网络技术进行第 2级粗分类 ,以减少候选字个数 ,提高识别速度 ;最后 ,利用汉字的笔划特征做最终细判识别 ,并给出识别结果汉字 .该系统的正确识别率达 98%以上 ,识别速度为 3字 s .该方法中结合了传统模式识别技术和神经网络技术 ,克服了不稳定性和不准确性 。
A kind of printed Chinese character recognition method is proposed and a corresponding system is implemented referring to 500 common used Chinese characters in table. firstly, preprocess the input Chinese character image and get the normalized Chinese character image. Then carry out the first level rough classification by using the area of Chinese character image and the second level rough classification by means of Artificial Neural Network technology. The purpose of doing so is to reduce the number of candidate Chinese character and raise the recognition speed. Finally, carry out the final fine discrimination by using the stroke of Chinese character and give out the resulting Chinese character. The recognition ratio of the system can get 98% and the recognition speed is 3 Chinese character per second. Because the conventional pattern recognition technology and Artificial Neural Network technology are combined, the instability and indeterminacy are avoided and the system is highly adaptable.
出处
《大庆石油学院学报》
CAS
北大核心
2002年第3期56-58,共3页
Journal of Daqing Petroleum Institute
关键词
表格处理
汉字识别方法
印刷体汉字
人工神经网络
printed Chinese character
Artificial Neural Network
Chinese character recognition
table processing
