摘要
本文从实际应用出发，在对预处理中存在的问题进行分析解决的基础上，开发了一套基于神经网络和锯法结构相结合的邮政编码自动识别系统。本系统软件采用Ｃ语言编写在４８６微机上实现。在对１３９７０个扫描数据测试中，其识别率为９８．４３％，产生的误识率为０．１３６％，拒识率为１．４３％。
From the point of application and based on the analysing of the problems,this paper introduced a zip code recognition system which combined the neural networks and the syntax structure method together. This software is written with languege C, it is runing on the 486 PC. The result of testing on 13970 digits minifests that the recognition rate is 98.43% and the refusal rate is 1.43% and the error rate is 0.136%.
出处
《中文信息学报》
CSCD
1995年第3期43-58,共16页
Journal of Chinese Information Processing
