基于框线检测的票据图像分类方法  

Bill Image Classification Method Based on Line Detection

作　　者：张艳[1] 张重阳[1] 郁生阳[2] 杨静宇[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,江苏南京210094 [2]上海交通大学图像处理与模式识别研究所,上海200240

出　　处：《南京理工大学学报》2007年第4期409-413,共5页Journal of Nanjing University of Science and Technology

基　　金：国家自然科学基金(60632050;60472060)

摘　　要：针对表格型票据,提出了一种基于框线检测的多类别票据图像分类方法。准确提取票据图像的框线特征,根据票据框线的拓扑结构建立起模板库;通过基于框线相关性的相似度模型,与标准模板库中模板的框线特征进行匹配,进而确定票据的分组类别。实际的票据分类实验证实了该方法的有效性和鲁棒性,为票据自动处理打下了坚实的基础。A new multi-sort bill classification algorithm based on line detection is proposed. The line feature of bill image sample, which is preparatory to be identified, is exactly detected according to the attribute of financial documents＇ structure. Based on the topological structure of lines, a standard template database is established reliably. In order to further confirm the category of bill group, template matching between sample and bill templates is put up through a new correlative matching model. The experimental results from real bill images demonstrate the efficiency and robustness of this method, which provides a stable foundation for automatic financial document processing.

关 键 词：票据分类 直线检测 相似性度量 框线匹配 

分 类 号：TP391.13[自动化与计算机技术—计算机应用技术]