基于Matlab和C#的数显仪表数字识别系统  

Digital Recognition System of Numerical Instruments Based on Matlab and C#

作　　者：张宁[1] 秦德鑫[2] 王秀芳[2] 

机构地区：[1]北京理工大学信息与电子学院,北京100081 [2]东北石油大学电气信息工程学院,黑龙江大庆163318

出　　处：《计算机技术与发展》2013年第9期70-73,共4页Computer Technology and Development

基　　金："十二五"国家科技支撑重点项目(2012BAH12B00);黑龙江省教育科学技术研究项目(12521056)

摘　　要：数显仪表中数字的识别技术在仪表自动识别领域中应用广泛,改善其识别准确率有助于提高仪表自动化水平。文中采用Otsu算法实现图像二值化,基于孤立像素连通域法对二值图像去噪,并利用垂直投影算法完成字符的分割,利用模块法进行特征提取。构建了3层BP神经网络,采用自适应带动量项的方法对BP神经网络进行参数调整。基于动态链接库方法,文中设计了结合Matlab和C#的数字识别系统。测试结果表明,单字符图像识别准确率可达98%,多字符图像识别准确率可达92.5%。Digital recognition technology is widely applied in numerical instruments. Improving its recognition rate is helpful for increasing the automation level of instruments. In this paper, Otsu algorithm is used to realize the image binarization ,isolated pixel connected domain is adopted to eliminate noise, vertical projection algorithm is applied to accomplish the character segmentation and modular algorithm is exploited to extract the feature. Furthermore,3-1ayered BP neural network is established,and the adaptive learning factor with a momen- tum term is used to adjust the parameters of BP neural network. Based on dynamic link library method ,the composite Maflab and C# dig- ital recognition system is designed. The results indicate that the recognition rate can reach 98% for single-number images and 92.5% for multi-number images.

关 键 词：数字识别 神经网络 特征提取 数显仪表 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]