一种基于BP网络的错帖检测新算法  

New algorithm of paper checking based on BP neural networks

作　　者：刘喆[1] 王跃宗[1] 杨建坤[1] 李德胜[1] 

机构地区：[1]北京工业大学机电学院,北京100022

出　　处：《计算机工程与应用》2007年第15期221-223,242,共4页Computer Engineering and Applications

摘　　要：提出了一种基于BP神经网络与图像不变矩的页面识别方法。应用规则矩的快速算法对一幅图像计算,将得到的7个图像矩不变量作为识别特征向量,输入到BP神经网络中进行识别,并最终确定出分类对象以达到对页面文字、图形以及图文的识别。通过实验仿真表明应用文中的方法能够对页面进行快速的训练与识别,既得到了较高的识别率,又满足了实时性要求。In this paper,a page-detecting method based on BP neural networks and moment invariants of pictures is proposed.It uses a fast algorithm to compute regular moments of a picture and at last gets seven moment invariants as a recognition vector. The moment invariant vector is put into BP neural networks to be recognized and classified,which is applied to characters and pictures.The experiment indicates that this method can train and recognize paper quickly.It not only gets high recognition rate but also satisfies real-time request.

关 键 词：BP神经网络 快速矩计算方法 页面检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]