查询结果:   李智杰,李昌华,刘欣,张沛,王玉英.融合拓扑特征和领域特征的非精确图匹配算法[J].计算机应用与软件,2015,32(10):164 - 167.
中文标题
融合拓扑特征和领域特征的非精确图匹配算法
发表栏目
人工智能与识别
摘要点击数
771
英文标题
INEXACT GRAPH MATCHING ALGORITHM INTEGRATING TOPOLOGICAL FEATURES AND DOMAIN FEATURES
作 者
李智杰 李昌华 刘欣 张沛 王玉英 Li Zhijie Li Changhua Liu Xin Zhang Pei Wang Yuying
作者单位
西安建筑科技大学信息与控制工程学院 陕西 西安 710055 西安建筑科技大学建筑学院 陕西 西安 710055 西安建筑科技大学理学院 陕西 西安 710055   
英文单位
College of Information and Control Engineering, Xi’an University of Architecture and Technology, Xi’an 710055,Shaanxi,China College of Architecture, Xi’an University of Architecture and Technology, Xi’an 710055,Shaanxi,China College of Science, Xi’an University of Architecture and Technology, Xi’an 710055,Shaanxi,China   
关键词
结构模式识别 空间句法 拓扑 统计模式识别 非精确图匹配
Keywords
Structural pattern recognition Space syntax Topology Statistical pattern recognition Inexact graph matching
基金项目
国家自然科学基金项目(61373112,51348002,50878176);陕西省教育厅专项科研项目(2013JK1157);西安建筑科技大学青年基金项目(QN1232)
作者资料
李智杰,博士生,主研领域:数字建筑,图形图像处理,模式识别。李昌华,教授。刘欣,博士生。张沛,教授。王玉英,副教授。 。
文章摘要
针对结构模式识别领域中现有图匹配算法对反映图本身拓扑结构的节点特征挖掘不够充分的问题,提出融合拓扑特征和领域特征的非精确图匹配算法。利用建筑学与城市规划学科中的空间句法理论构造图拓扑特征的量化描述,并将其与节点属性和边属性等其他领域的非拓扑特征相结合,构造描述图特征的特征向量,以此为桥梁将结构模式识别问题转化为统计模式识别问题,进而借助支持向量机实现非精确图匹配。不同于其他的图匹配算法,该算法对图的拓扑表达能力强,并且可融合图的领域方面的非拓扑特征,通用性较好。实验结果表明,提出的图匹配算法在不同的图数据集上均具有较高的分类识别率。
Abstract
In the field of structural pattern recognition, the existing graph matching algorithms can’t efficiently mine the node features reflecting the topological structures of graph itself. To solve this problem, we propose a new inexact graph matching algorithm which integrates the topological features and domain features. We use the space syntax theory in architecture and urban planning to construct the quantitative description of graph’s topological features, and then combine them with non-topological features in other domain aspects, such as node attributes and edge attributes, etc., to construct the feature vectors which depict the graph feature. In this way, the structural pattern recognition is converted to statistical pattern recognition, and the SVM can then be used as the aid to achieve inexact graph matching. Differing from other graph matching methods, the proposed algorithm can adequately render the graph’s topology and merge the non-topological features in terms of the graph’s domain property, and has a favourable universality as well. Experimental results show that the proposed graph matching algorithm can achieve higher classifying accuracy in different graph datasets.
下载PDF全文