集成汉英OCR系统识别中文名片

汉英双语混排识别是构造中文自动文档图像处理系统时常会遇到的一个问题。只有采用一种有效的方法集成现有汉英识别引擎,才可能高质量地识别混排文档。该文应用适当干预和多层次语言判断的汉英OCR系统集成原则,集成OCR系统识别中文名片。实验数据表明,利用该原则构造的系统确实能有效集成汉英识别引擎,在纯中文识别率为89.86%,纯英文识别率为91.20%的情况下,使名片最终总体识别率达到了93.45%,较好地解决了汉英混排名片的识别问题。

Integrating Chinese and English OCR to Recognize Chinese Business Card

Chinese and English bilingual OCR is a common problem to be resolved in Automatic Chinese Document Imaging Management (ACDIM).To achieve high performance recognition in a language mixed environment ,people need an effective way to integrate and to make use of two existing engines,one for Chinese OCR and the other for English OCR.This paper proposes an integration scheme to Chinese Business Card recognition based on″Proper Adjustment″and″Multi-layer Language Transition″.A real system experiment shows the successfulness of integration,which brings the system recognition rate to93.45%,higher than the original accuracy of each lower lever engine(pure Chinese OCR:89.86%,pure English OCR:91.20%).The mixed language recognition problem is mostly resolved in this system.