摘要
民国纸币的冠字号码图像分割是民国纸币自动识别系统的关键组成部分。民国纸币种类繁多,不同纸币背景图案、字体、主色调差异较大且绝大多数纸币因保存时间较长,存在褶皱、破损、污迹等情况,冠字号码提取与分割较为困难。文章分析了民国时期不同银行、不同年代、不同地域纸币的特点,提出一种民国纸币冠字号码提取与分割算法。首先利用颜色特征、形态学处理、连通域形状测量提取冠字号码,然后使用基于逐差与滑动窗口的字符分割算法分割单个字符。实验结果表明,所提算法对民国纸币冠字号码有很好的提取与分割效果,分别取得了99.3%、98.5%的准确率,为后续字符识别奠定了良好基础。
Article number image segmentation is a key part of automatic recognition system of banknotes of the republic of China. There were numerous kinds of banknotes with different background patterns, fonts and main colors, and most of the banknotes were folded, damaged or stained due to the long preservation time, so it was difficult to extract and segment the article numbers. This article analyzed the characteristics of banknotes in different banks, different years, and different regions during the period of Republic of China, and proposed an algorithm for extracting and segmenting the prefix of the banknotes of the Republic of China. Firstly, we extracted article numbers by using color features, morphological processing and connected domain shape measurements, then adopted character segmentation algorithm based on step by step difference and Sliding window.The experimental results show that the algorithm presented in this paper has a good extraction and segmentation effect on the article number of banknotes of the republic of China with an accuracy of 99.3% and 98.5% respectively, laying a good foundation for the subsequent character recognition.
作者
沈成龙
王笑梅
王晨
SHEN Cheng-long;WANG Xiao-mei;WANG Chen(Shanghai Normal University,Shanghai 200030,China)
出处
《计算机仿真》
北大核心
2022年第1期437-440,455,共5页
Computer Simulation
基金
Shanghai Science and Technology Innovation Fund(17060502600)。
关键词
民国纸币
冠字号码
颜色特征
形态学
图像处理
Banknotes of the Republic of China
Article number
Color characteristics
Morphology
Image processing
