摘要
主要研究了票据图像处理中票据类型识别问题.该算法分为特征线提取和版面匹配两部分.在分析以往图像中直线提取算法之优缺点的基础上,根据票据图像中线条的特点给出了一个适用于票据图像的快速线段检取算法,其时间复杂度为O(n).在特征线检取的基础上,根据不同票据有不同几何和拓扑结构这一原理,给出了一种基于特征线检取的票据类型识别算法,并在Windows环境下用VisualC++实现了上述算法.实验结果表明,该算法用于票据识别速度快且具有较高的识别率和很强的实用性.
Among all the automatic financial document processi ng steps, financial document recognition is in the first place. This paper describ es algorithms currently under development of financial document recognition base d on characteristic line detection. It consists of two major parts. A l ine detection algorithm applied in financial document images is first proposed a ccordi ng to the specific characters of financial documents' structure. Then, a fin ancial document pattern recognition algorithm based on the fact that different f inancial documents have different geometrical structures is also proposed. At th e end of this paper, some experimental results will be listed. The experimental results showed that the algorithms are fast, reliable a nd very practical.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2003年第2期173-177,共5页
Journal of Zhejiang University：Engineering Science
关键词
特征抽取
特征线
票据特征库
原点
匹配确信度
character retrieval
characteristic lines
check character library
origin
match threshold
