印章自动鉴定系统中的印章图像检索技术研究  

Research on Seal Imprint Image Retrieval Technique in Automatic Seal Imprint Verification System

作　　者：崔云凡[1] 梁阿磊[1] 

机构地区：[1]上海交通大学计算机科学与工程系

出　　处：《微型电脑应用》2013年第8期27-30,共4页Microcomputer Applications

摘　　要：在自动印章鉴定的过程中,需要利用待测印章的图像从印章数据库中检索出作为对照的标准印章图像。印章图像检索同其他图像检索一样,需要解决语义鸿沟的问题,而印章图案的残缺又给这一问题带来了困难。在对现有基于内容的图像检索(Content-based Image Retrieval,CBIR)技术基础上,提出了基于深度信念网络(Deep Belief Network,DBN)的印章图像检索方案,利用DBN来弥补印章图像检索中的语义鸿沟,并设计了相应的数据以适应该检索方法。实践表明,利用该方法,能够有效地利用待测印章图像从数据库中检索出对应的标准印章图像。In automatic seal imprint （or seal） verification, the seal image to be verified is used, as a query, to retrieve the standard seal image from seal database. As regular image retrieval, seal image retrievel is also faced with sematic gap, and the incomplete seal patterns make it more difficult to achieve. Based on current Content-based Image Retrieval （CBIR） techniques, a new scheme which uses Deep Belief Network （DBN） to reduce sematic gap is proposed, and a particularly designed database system is implemented to adapt to DBN retrieval method in this paper. The practice shows that using our method, the standard seal image can be effectively retrieved with a query of seal image to be verified.

关 键 词：印章检索 语义鸿沟 深度信念网络 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]