第14卷第2期 智能系统学报 Vol.14 No.2 2019年3月 CAAI Transactions on Intelligent Systems Mar.2019 D0:10.11992/tis.201709038 网络出版地址：http:/kns.cnki.net/kcms/detail/23.1538.tp.20180417.1720.010.html 基于图表示和匹配的表单定位与提取 谭婷1，吕淑静2，吕岳2 (1.华东师范大学上海多维度信息处理重点实验室，上海200062：2.中国邮政集团公司上海研究院图像分析 与智能系统联合实验室，上海200062) 摘要：为了实现对不同类型、分辨率和方向的快递表单上用户感兴趣区域信息的获取，本文提出了一种基于 图表示和匹配的表单定位与提取方法。选择参考表单中已有的印刷图案或字符等关键区域作为基准位置，进 行图的表示。基于图像分割得到的候选关键区域对待处理表单进行图表示。然后，根据图的属性计算待处理 表单与参考表单的相似度。最后，将最大相似度对应的同构图作为参考表单图的最优匹配，并建立同构图与参 考表单图位置映射，定位出表单。本文实验数据集来源于真实场景下采集的快递包裹表单图像。实验结果表 明：本文算法在快递包裹表单图像上具有良好的性能，对旋转、光照变化、局部遮挡具有较好的鲁棒性。 关键词：图像分割：表单提取：表单定位：图表示：图匹配：同构图：快递包裹分栋 中图分类号：TP751.1文献标志码：A文章编号：1673-4785(2019)02-0231-08 中文引用格式：谭婷，吕淑静，吕岳.基于图表示和匹配的表单定位与提取J.智能系统学报，2019,14(2)：231-238. 英文引用格式：TAN Ting,LYU Shujing,LYU Yue..Form location and extraction based on graph representation and matching Jl, CAAI transactions on intelligent systems,2019,14(2):231-238. Form location and extraction based on graph representation and matching TAN Ting',LYU Shujing,LYU Yue2 (1.Shanghai Key Laboratory of Multidimensional Information Processing,East China Normal University,Shanghai 200062,China; 2.ECNU-SRI Joint Lab for Pattern Analysis and Intelligent System,Shanghai Research Institute of China Post,Shanghai 200062. China) Abstract:To obtain information of a user's interested region on express package images of different types,resolutions, and directions,a form location and extraction method based on graph representation and matching is proposed in this pa- per.A reference form is needed in this method.First,key regions such as the existing printed patterns or characters in the reference form are chosen as nodes to build the reference graph.Second,graph representation is conducted on the form to be processed based on the candidate key region derived from image segmentation.Then,the similarity between the reference form and the candidate form is calculated according to attributes of the graph.Finally,the isomorphic graph with the maximum similarity is chosen as the optimal matching of the reference form and graph,and the position mapping of the isomorphic graph and the reference form and test image is established to locate the form.The experi- mental datasets in this paper originate from express package images collected in practical scenarios.Experimental res- ults indicate that the proposed algorithm has good performance on express form images.Especially,good robustness is achieved for rotated,illuminated,and partially shaded images. Keywords:image segmentation;form extraction;form location;graph representation;graph matching;isomorphic graph;express package sorting 表单作为重要的信息载体在实际生活和工作 符号等都有可能包含用户感兴趣的重要信息，如 中有着广泛的运用，表单中某些特定字段、图案、 订货单中的订单号、发票的具体项目及金额、快 收稿日期：2017-09-20.网络出版日期：2018-04-18. 递运单中的收货地址和手机号码等。人工录入的 通信作者：谭婷.E-mail:tanting_.hn@l63.com. 方式采集数据，费时费力，而且容易出错，因此利DOI: 10.11992/tis.201709038 网络出版地址: http://kns.cnki.net/kcms/detail/23.1538.tp.20180417.1720.010.html 基于图表示和匹配的表单定位与提取 谭婷1，吕淑静2，吕岳1,2 （1. 华东师范大学 上海多维度信息处理重点实验室，上海 200062; 2. 中国邮政集团公司上海研究院 图像分析 与智能系统联合实验室，上海 200062） 摘 要：为了实现对不同类型、分辨率和方向的快递表单上用户感兴趣区域信息的获取，本文提出了一种基于 图表示和匹配的表单定位与提取方法。选择参考表单中已有的印刷图案或字符等关键区域作为基准位置，进 行图的表示。基于图像分割得到的候选关键区域对待处理表单进行图表示。然后，根据图的属性计算待处理 表单与参考表单的相似度。最后，将最大相似度对应的同构图作为参考表单图的最优匹配，并建立同构图与参 考表单图位置映射，定位出表单。本文实验数据集来源于真实场景下采集的快递包裹表单图像。实验结果表 明：本文算法在快递包裹表单图像上具有良好的性能，对旋转、光照变化、局部遮挡具有较好的鲁棒性。 关键词：图像分割；表单提取；表单定位；图表示；图匹配；同构图；快递包裹分拣 中图分类号：TP751.1 文献标志码：A 文章编号：1673−4785(2019)02−0231−08 中文引用格式：谭婷, 吕淑静, 吕岳. 基于图表示和匹配的表单定位与提取 [J]. 智能系统学报, 2019, 14(2): 231–238. 英文引用格式：TAN Ting, LYU Shujing, LYU Yue. Form location and extraction based on graph representation and matching[J]. CAAI transactions on intelligent systems, 2019, 14(2): 231–238. Form location and extraction based on graph representation and matching TAN Ting1 ，LYU Shujing2 ，LYU Yue1,2 (1. Shanghai Key Laboratory of Multidimensional Information Processing, East China Normal University, Shanghai 200062, China; 2. ECNU-SRI Joint Lab for Pattern Analysis and Intelligent System, Shanghai Research Institute of China Post, Shanghai 200062, China) Abstract: To obtain information of a user’s interested region on express package images of different types, resolutions, and directions, a form location and extraction method based on graph representation and matching is proposed in this pa￾per. A reference form is needed in this method. First, key regions such as the existing printed patterns or characters in the reference form are chosen as nodes to build the reference graph. Second, graph representation is conducted on the form to be processed based on the candidate key region derived from image segmentation. Then, the similarity between the reference form and the candidate form is calculated according to attributes of the graph. Finally, the isomorphic graph with the maximum similarity is chosen as the optimal matching of the reference form and graph, and the position mapping of the isomorphic graph and the reference form and test image is established to locate the form. The experi￾mental datasets in this paper originate from express package images collected in practical scenarios. Experimental res￾ults indicate that the proposed algorithm has good performance on express form images. Especially, good robustness is achieved for rotated, illuminated, and partially shaded images. Keywords: image segmentation; form extraction; form location; graph representation; graph matching; isomorphic graph; express package sorting 表单作为重要的信息载体在实际生活和工作 中有着广泛的运用，表单中某些特定字段、图案、 符号等都有可能包含用户感兴趣的重要信息，如 订货单中的订单号、发票的具体项目及金额、快 递运单中的收货地址和手机号码等。人工录入的 方式采集数据，费时费力，而且容易出错，因此利 收稿日期：2017−09−20. 网络出版日期：2018−04−18. 通信作者：谭婷. E-mail：tanting_hn@163.com. 第 14 卷第 2 期 智 能 系 统 学 报 Vol.14 No.2 2019 年 3 月 CAAI Transactions on Intelligent Systems Mar. 2019