基于BP神经网络的实时水表自动识别系统的研究  

Research on Real-Time Water Meter Recognition System Based on BP Neural Network

作　　者：李云栋[1] 孙长庆[1] 

机构地区：[1]北方工业大学信息工程学院,北京100144

出　　处：《计算机测量与控制》2013年第5期1330-1332,共3页Computer Measurement &Control

摘　　要：为提高数字水表字符的识别率,提出一种基于数字图像处理技术与BP神经网络的水表自动识别系统;首先对采集的水表图像进行预处理以获得字符区域,然后进行二值化操作,利用连通区域提取算法分割出单个字符,最后利用训练好的BP神经网络对分割后的字符进行识别,并输出识别结果;实际运行表明自动识别系统正确率在95%以上,并且能够实时显示水表数据。In order to improve the recognition rate of the digital water meter characters, a water meter automatic recognition system based on image processing and BP neural network is proposed in this paper. Firstly, characters area is located through pre--processing water meter image. Secondly, single character is segmented by connection component algorithm after binary thresholding. Finally, recognition re- sult is obtained by pre--trained BP neural network system. Experimental results show the accuracy of the proposed automatic recognition system more than 95%, and it can real time display meter data.

关 键 词：数字水表 字符识别 BP神经网络 连通区域提取 字符分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]