一种数显仪表数字字符识别方法研究  

Research on a character recognition method for numerical instruments

作　　者：巩玉滨[1] 杨红娟[2] 张运楚[2] 常国雷[1] 邱冬[1] 

机构地区：[1]山东信息通信技术研究院管理中心,山东济南250101 [2]山东建筑大学信息与电气工程学院,山东济南250101

出　　处：《山东建筑大学学报》2011年第2期134-137,177,共5页Journal of Shandong Jianzhu University

基　　金：山东省自然科学基金项目(2006F12)

摘　　要：针对工业控制领域数显仪表检测系统数据自动采集中数显仪表图像存在的歪斜、缺损和模糊问题,给出了一种有效的数显仪表数字字符识别方法。通过Radon变换找到倾斜角度纠斜校正图像;采用腐蚀细化修补图像的缺损;运用行分割、单字分割与规范化进行数字字符的准确分割模糊图像,以网格法与交线特征提取法提取数显仪表数字字符的特征,采用三层BP神经网络进行数字字符识别。某煤矿企业实际采集的数显仪表彩色图像数字字符识别实例,说明本文方法识别率高,抗干扰性好。In view of the inclination, defection and unclearness, a practical character recognition method for numerical instruments is studied to collect data automatically from numerical instrument detection system in industrial control field. The inclination angle of image is fixed and adjusted through coordinate rotating transition method Radon transformation. Image corrosion and thinning are used to handle defects and improve image quality. Row segment, single character segment and normalization are used to segment precisely the numerical characters from blurred images. The numerical character feature is extracted with grid method and intersection feature extraction method. Three-layer BP neural network is selected to recognize numerical characters. The numerical characters recognition example of a coal company numerical instrument image shows that the proposed method has a high recognition rate and a good anti-interference.

关 键 词：数显仪表 字符识别 RADON变换 BP神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]