[1]朱 彬,薛路强,谭守标.一种改进的数字图像定位识别方法研究[J].计算机技术与发展,2017,27(12):67-70.[doi:10.3969/ j. issn.1673-629X.2017.12.015]
 ZHU Bin,XUE Lu-qiang,TAN Shou-biao.Research on an Improved Auto-locating and Recognition Method for Digital Images[J].Computer Technology and Development,2017,27(12):67-70.[doi:10.3969/ j. issn.1673-629X.2017.12.015]
点击复制

一种改进的数字图像定位识别方法研究()
分享到:

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
27
期数:
2017年12期
页码:
67-70
栏目:
智能、算法、系统工程
出版日期:
2017-12-10

文章信息/Info

Title:
Research on an Improved Auto-locating and Recognition Method for Digital Images
文章编号:
1673-629X(2017)12-0067-04
作者:
朱 彬1 薛路强2 谭守标2
1. 国网安庆供电公司 安全监察质量部,安徽 安庆 246000;
2. 安徽大学 计算智能与信号处理教育部重点实验室,安徽 合肥 230039
Author(s):
ZHU Bin 1 XUE Lu-qiang 2 TAN Shou-biao 2
1. Department of Safety Supervision Quality,State Grid Anqing Electric Power Supply Co. ,Anqing 246000,China;
2. Key Lab of Intelligent Computing &Signal Processing of Ministry of Education,Anhui University,Hefei 230039,China
关键词:
笔画宽度变换算法读数精确定位多层次扩展合并读数识别
Keywords:
stroke width transformdigits auto-locatingmulti-level extension and mergingreading recognition
分类号:
TP391.41
DOI:
10.3969/ j. issn.1673-629X.2017.12.015
文献标志码:
A
摘要:
针对图像视频中数字自动识别处理的需求,提出了一种改进的数字区域定位及读数识别方法。 该方法使用自适应阈值进行图像整体二值化,然后设计改进的笔画宽度变化算法(SWT)来确定仪表数字显示的大体位置,再根据数字的颜色、宽高比以及空间排列等特征来过滤得到准确位置,并使用多层次扩展合并处理方法去除遮挡粘连影响,实现读数区域的精确定位,效果理想。 最后对数字区域提取多种高区分度特征,通过训练好的多分类模型即可准确识别得到对应数字值,实现图像视频中读数的自动识别。 实验结果表明,该方法具有很高的准确度及较强的鲁棒性,能避免光照、倾斜、部分遮挡的影响,准确找到读数区域,并据此识别出其中的数字,适用于自动巡检、远程抄表等多种应用。
Abstract:
According to the requirement of the automatic recognition for digital video,an improved digits auto-locating and recognition method is presented. It adopts self-adaptive threshold for binarization of image and then an proved algorithm of Stroke Width Transform (SWT) is designed to make a coarse locating of the digits’ regions. After that,the precise positions of the digits are determined by filtering them with some useful features,such as its height-width-ratio,color and spatial arrangement,and the multi-level extension and merging is applied to liminate the influence on shield and adhesion for the exact locating of digits region with perfection. At last,after extraction of the high discriminative features in digital regions,the digits can be accurately recognized and achieved by trained multi-classified models,which can implement the automatic recognition of digits in videos. The experimental results show that the proposed method owns high accuracy and strong robustness,without impact on light,titlt and partial shield,and locate the correct digits regions for recognition of digits. It is suitable for automatic inspection,remote meter reading and so on.
更新日期/Last Update: 2018-03-06