摘要
本文介绍了一种实用的字符识别方法。采用CCD摄像机摄取仪表读数,计算机处理系统在提取图像数字特征的基础上完成对字符的自动识别,并快速、准确地将结果进行显示。采用多帧差异积累方法实现了数字区域的自动定位分割;采用高斯滤波、Canny边缘检测和Hough变换的方法校正字符的倾斜角度;采用自适应阈值、形态滤波算法去除因图像亮度不均匀对字符识别准确度的影响;采用纵向投影方法进行字符识别。在现场试验中,取得了令人满意的效果。
A method available for digital character recognition is introduced in this paper. Using a CCD camera to take the images of a digital meter, the digital characters are partitioned and recognized accurately and displayed in real-time by a computer system. This paper expatiates the algorithms of character recognition of the digital meters including the methods of digital image processing based on multi-frame difference accumulation, which located digital area, Gauss filtering, Canny edge detection, Hough transformation which calibrated slant angle, the adaptive threshold, which are used to position the digital characters, and vertical projection which recognized the digitals. A lot of experiments applied in real verified and the correct distinguish results for the digital characters are satisfied.
出处
《微计算机信息》
北大核心
2008年第28期219-221,共3页
Control & Automation
基金
吉林省杰出青年资助基金(20060119)
颁发部门:吉林省科技厅
关键词
视频图像
字符识别
自适应阈值
数字区域定位
HOUGH变换
video inaage
character recognition
adaptive threshold
digital characters positioning
Hough transformation
作者简介
尹传历（1979．2-），中国科学院长春光学精密机械与物理研究所博士研究生，主要研究方向为模式识别、视频图像处理技术等；
宋建中（1942．4-），中国科学院长春光学精密机械与物理研究所博士生导师，研究员，主要研究方向为模式识别与电视跟踪技术。
