摘要
本文介绍了采用综合技术集成的方法,解决印刷体汉字识别系统误识率太高的重大难题,并通过集成系统的实践,证实了其技术集成优势.由于识别方法的互补效应,不仅提高了识别的正确率,而且使误识率得到大幅度的降低.采用该集成办法研制的系统,经过100万字的实际文章的测试,系统的识别率超过98%,误识率小于0.2%,尤其是汉字的误识率小于0.1%.
High error rate is an essential problem for an OCR system. It is diffcult to develop an OCR system with both high correct rate and low error rate. This paper introduces an integrated system for printed Chinese character recognition. It integrates three recognition methods into an OCR system. Because the three methods are complemetary each other,the integrated system achieved high correct rate and very low error rate. A test about 1000000 printed Chinese characters shows that the correct rate is above 98% and the error rate is less than 0. 2%,especially the Chinese character error rate is less than 0. 1 %.
出处
《计算机学报》
EI
CSCD
北大核心
1995年第9期678-685,共8页
Chinese Journal of Computers
基金
863高技术基金
关键词
模式识别
技术综合集成
汉字识别
Image processing, feature extraction,Chinese character recognition