Feature Matching via Topology-Aware Graph Interaction Model

Feature matching plays a key role in computer vision. However, due to the limitations of the descriptors, the putative matches are inevitably contaminated by massive outliers.This paper attempts to tackle the outlier filtering problem from two aspects. First, a robust and efficient graph interaction model,is proposed, with the assumption that matches are correlated with each other rather than independently distributed. To this end, we construct a graph based on the local relationships of matches and formulate the outlier filtering task as a binary labeling energy minimization problem, where the pairwise term encodes the interaction between matches. We further show that this formulation can be solved globally by graph cut algorithm. Our new formulation always improves the performance of previous localitybased method without noticeable deterioration in processing time,adding a few milliseconds. Second, to construct a better graph structure, a robust and geometrically meaningful topology-aware relationship is developed to capture the topology relationship between matches. The two components in sum lead to topology interaction matching(TIM), an effective and efficient method for outlier filtering. Extensive experiments on several large and diverse datasets for multiple vision tasks including general feature matching, as well as relative pose estimation, homography and fundamental matrix estimation, loop-closure detection, and multi-modal image matching, demonstrate that our TIM is more competitive than current state-of-the-art methods, in terms of generality, efficiency, and effectiveness. The source code is publicly available at http://github.com/YifanLu2000/TIM.

Topology and Semantic Information Fusion Classification Network Based on Hyperspectral Images of Chinese Herbs

Most methods for classifying hyperspectral data only consider the local spatial relation-ship among samples,ignoring the important non-local topological relationship.However,the non-local topological relationship is better at representing the structure of hyperspectral data.This paper proposes a deep learning model called Topology and semantic information fusion classification network(TSFnet)that incorporates a topology structure and semantic information transmis-sion network to accurately classify traditional Chinese medicine in hyperspectral images.TSFnet uses a convolutional neural network(CNN)to extract features and a graph convolution network(GCN)to capture potential topological relationships among different types of Chinese herbal medicines.The results show that TSFnet outperforms other state-of-the-art deep learning classification algorithms in two different scenarios of herbal medicine datasets.Additionally,the proposed TSFnet model is lightweight and can be easily deployed for mobile herbal medicine classification.

Schultz Polynomials and Their Topological Indices of Jahangir Graphs J2,m

Let G = (V;E) be a simple connected graph. The Wiener index is the sum of distances between all pairs of vertices of a connected graph. The Schultz topological index is equal to and the Modified Schultz topological index is . In this paper, the Schultz, Modified Schultz polynomials and their topological indices of Jahangir graphs J2,m for all integer number m ≥ 3 are calculated.

基于多维动态拓扑学习图卷积的骨架动作识别

图卷积由于其对图数据的强大表示能力被广泛应用于基于骨架的动作识别任务中.但是现有的图卷积方法在所有帧或通道上都使用共享的图拓扑进行特征聚合,这极大限制了图卷积网络的表示能力.为了解决这些问题,本文提出多维动态拓扑学习图卷积用于动态建模具有时序与通道特异性的拓扑结构.多维动态拓扑学习图卷积主要包含三个组成部分:纯粹节点拓扑学习图卷积(pure Joint topology learning Graph Convolution,J-GC)、动态时序特异性拓扑学习图卷积(Dynamic Temporal-Wise topology learning Graph Convolution,DTW-GC)和通道特异性拓扑学习图卷积(Channel-Wise topology learning Graph Convolution,CW-GC).特别地,在DTW-GC中使用了动态骨架拓扑建模方法(Dynamic Skeleton Topology Learning,DSTL),以高效地建模富含全局时空拓扑特征的动态骨架拓扑.将多维动态拓扑学习图卷积与多尺度时间卷积(Multi-Scale Temporal Convolution,MS-TC)相结合,本文构建了具有强大建模能力的图卷积网络.此外,为了对骨架数据的空间信息进行补充,本文额外引入了相对节点数据和相对骨骼数据进行多流网络的融合.本文所提出的方法在NTU-RGB+D与NTU-RGB+D 120数据集上分别取得了92.64%和89.29%的准确率,超过了当前最先进方法.

基于图论的动态调整拓扑限制短路电流措施对系统可靠性影响研究

随着电力系统规模的逐渐扩大和负荷密度的提升,短路电流超标问题逐渐成为制约电网发展的重要因素。基于系统拓扑动态调整的短路电流限制措施,不影响系统正常运行,仅在故障期间调整系统结构,这一措施的引入可能对系统的可靠性造成一定的影响。该文研究了电网连通性的图论建模方法及计算方法,并以典型电网为例建立了系统拓扑仿真算例,分析了不同拓扑调整方法下,短路电流抑制效果和对系统可靠性的影响。结果表明,220kV系统典型接线下,拓扑调整开关配置母线分段开关处可作为首先考虑的方案,其短路电流抑制效果较好,对系统可靠性影响相对较小。

电动汽车锂电池组双层环形均衡器研究

传统Buck-Boost均衡电路传输路径长且均衡效率低,现有的环形均衡器无法从根本上解决该问题。针对该问题,文中提出了一种基于Buck-Boost及开关电容的锂电池双层环形均衡器。该均衡电路采用模块化均衡的方法在底层模块和顶层模块采用Buck-Boost和开关电容的混合电路实现均衡,并采用图论分析的方法,从均衡速度、均衡效率两方面分析电路结构。结果表明,该均衡电路具有能量传输路径少、能量传输效率高的优点。文中在MATLAB/Simulink电力仿真平台上搭建了均衡电路的仿真模型,仿真结果表明基于Buck-Boost及开关电容的双层环形均衡器比单层环形均衡器均衡速度提高了46%。文中搭建了4节电池的实验样机对环形均衡电路工作原理及系统均衡效率进行实验验证,结果表明双层环形均衡器比单层环形均衡器均衡效率可提高26.78%。仿真与实验结果与理论分析一致,说明电路结构合理有效。

异质图嵌入的地理不敏感时空兴趣点推荐方法

基于地理位置的社交网络(LBSN)规模日渐庞大,促进了兴趣点(POI)推荐业务快速发展。现有研究常直接引入POI地理空间距离难以模拟用户高度随机的行为路径,导致兴趣点推荐过程对地点位置距离度量较为敏感。同时,社交网络中用户稀疏的POI签到数据也容易对推荐精度产生巨大影响。针对以上问题,提出一种基于异质图嵌入的地理不敏感时空兴趣点推荐模型(GIPR)。首先,引入用户行为序列,构造行为POI时空拓扑图,使用权重空间路径表示相对位置距离,既符合用户行为特征,也降低了推荐过程对兴趣点间距离的敏感程度,进而增强推荐结果的可解释性。其次,面对异质且高度稀疏的交互数据,提出的GIPR推荐方法可从局部和全局对完整的LBSN异质图进行学习,融合更丰富的用户和POI特征。最后,经过注意力层提取用户的长短期偏好,实现更加个性化的兴趣点推荐。在两个大规模真实数据集Foursquare和Gowalla上的实验表明,GIPR方法具有更高的推荐精度与更强的可解释性。

融合拓扑势与因子图的在线社交网用户影响力推断

拓扑势可用于计算在线社交网用户影响力,但是,其对所有用户优化同一个影响参数,导致影响力以相同速度衰减的缺陷尚待改进.此外,拓扑势只适用于节点质量相同的网络.对此,本文融合拓扑势和因子图,提出影响力因子图(Impact Factor Graph,IFG)模型.IFG模型可推断网络用户节点间影响力,并对每个节点分配不同影响参数.维基百科合作编辑数据集上的实证表明,IFG模型可以解决拓扑势在异质网络上的计算问题,并提高用户影响力分析的合理性.

利用图模型存储算法依赖关系的方法

在大数据时代,用于数据处理的算法数量呈爆发式增长,当前对大量算法的管理方法通常是对算法分类、打标签或以任务为单位存储由算法构成的流程,对任务集合中的算法间拓扑关系未能给予足够的重视.随着领域知识与任务流程的积累,算法间的依赖关系愈发重要.本文基于巨量算法管理的需求,提出了拆分有分支依赖关系为无分支依赖关系的管理方法,通过免索引邻接图数据库的指针搜寻拓扑关系,避免Join操作,在管理算法依赖关系时具有先天优势.另外为突出算法模块复用能力,提出“连接点”的概念,在图模型中用节点表示依赖关系边,区分算法模块在不同任务流程的位置,使被多个任务复用的算法模块在图中只需用一个算法模块节点表示.最后,基于具体项目验证了本文提出的算法关系管理方法,证明本文算法关系管理方法在算法数量成规模且算法模块高复用的场景下具有明显优势.

Exponential networked synchronization of master-slave chaotic systems with timevarying communication topologies

The networked synchronization problem of a class of master-slave chaotic systems with time-varying communication topologies is investigated in this paper. Based on algebraic graph theory and matrix theory, a simple linear state feedback controller is designed to synchronize the master chaotic system and the slave chaotic systems with a time- varying communication topology connection. The exponential stability of the closed-loop networked synchronization error system is guaranteed by applying Lyapunov stability theory. The derived novel criteria are in the form of linear matrix inequalities （LMIs）, which are easy to examine and tremendously reduce the computation burden from the feedback matrices. This paper provides an alternative networked secure communication scheme which can be extended conveniently. An illustrative example is given to demonstrate the effectiveness of the proposed networked synchronization method.

A New Topological Description Method of Kinematic Chain

This paper presents a novel method for the description of kinematic chains, namely the canonical description of kinematic chains including the synthetic degree-sequences and the canonical adjacency matrices sets of kinematic chains. The most important characteristic of this new description method is its uniqueness. Based on the new principle the isomorphism identification becomes easy and the structures of all kinds of kinematic chains can be stored in computer for the benefits of the realization of automation and intelligence of machine design.

基于RedisGraph的电网CIM/XML数据模型构建与应用

电网拓扑分析在电力系统运行规划中应用广泛。现有研究中,电网拓扑信息多以关系型数据库或图数据库的方式存储于磁盘,在查询速度上受限于数据库本身的特性。为了提高电网拓扑信息的查询速度,文章提出了一种基于内存型图数据库RedisGraph的电网拓扑模型构建与查询方法。首先概述了CIM/XML文件的数据结构和RedisGraph存储数据的方式,然后据此设计了将CIM/XML数据模型转换为电网拓扑图数据模型的方案,并实现了基于RedisGraph的拓扑搜索,最后基于构建的图数据模型进行拓扑搜索测试。结果表明,提出的基于RedisGraph的电网拓扑构建与查询方案能更好地满足大规模复杂电网的拓扑分析业务的性能需求。

POINT PATTERN MATCHING ALGORITHM BASED ON POINT PAIR TOPOLOGICAL CHARACTERISTICS AND SPECTRAL MATCHING

Most of the Point Pattern Matching (PPM) algorithm performs poorly when the noise of the point's position and outliers exist. This paper presents a novel and robust PPM algorithm which combined Point Pair Topological Characteristics (PPTC) and Spectral Matching (SM) together to solve the afore mentioned issues. In which PPTC, a new shape descriptor, is firstly proposed. A new comparability measurement based on PPTC is defined as the matching probability. Finally, the correct matching results are achieved by the spectral matching method. The synthetic data experiments show its robustness by comparing with the other state-of-art algorithms and the real world data experiments show its effectiveness.

Some Topological Indices Computing Results If Archimedean Lattices L(4,6,12)

The introduction of graph-theoretical structure descriptors represents an important step forward in the research of predictive models in chemistry and falls within the lines of the increasing use of mathematical and computational methods in contemporary chemistry.The basis for these models is the study of the quantitative structure-property and structure-activity relationship.In this paper,we investigate Great rhom-bitrihexagonal which is a kind of dodecagon honeycomb net-work covered by quadrangle and hexagon.Many topological indexes of Great rhom-bitrihexagonal have being investigated,such as sum-connectivity index,atom-bond connectivity index,geometric-arithmetic index,fifth,harmonic index,Randićconnectivity index,first Zagreb index,second Zagreb index and the corresponding Zagreb polynomials,modified Zagreb index,fourth atom-bond connectivity index,augmented Zagreb index,hyper-Zagreb index,Sankruti index,forgotten topological index,first multiple Zagreb index,second multiple Zagreb index,as well as derived geometric-arithmetic index,Narumi-Katayama index and modified Narumi-Katayama index.

Computing Topological Invariants of Triangular Chandelier Lattice

A numerical parameter mathematically derived from the graph structure is a topological index.The topological index is the first actual choice in QSAR research and these indices are used to build the correlation model between the chemical structures of various chemicals compounds.Here,we investigate some old degree-based topological indices like Randic index,sum connectivity index,ABC index,GA index,1st and 2nd Zagreb indices,modified second Zagreb index,redefined version of 1st,2nd and 3rd Zagreb indices,hyper and augmented Zagreb indices,forgotten index and symmetric division degree index,and some new degree-based indices like SK index,SK1 index,SK2 index,and AG1 index of triangular chandelier-lattice(TCL).The results are generalized by using edge partition and closed formulas for topological indices of triangular chandelier-lattice are analysed.

Topological Expression Model of Satellite Gear Mechanism

By investigation of the topological characteristics of the kinematic structure of Satellite Gear Mechanism (SGM) with graph theory, the graph model of SGM is analyzed, and a topological expression model between input and output of SGM is established based on systematic design point. Meanwhile, the mathematical expression for SGM is deduced by integrating matrix theory and graph theory; thus, the topological characteristics of the kinematic structure of SGM can be converted into a matrix model, and the topological design problem of SGM into a matrix operation problem. In addition, a brief discussion about the measures for identification of isomorphism of the graph mode is made.

Basic Concepts of Graph Theory as the Instruments of Mathematical Formalization of Woven Structures

The paper presents the prerequisites of involving of topological elements and graph theory as an instrument of mathematical formalization of woven structures and technology of textile fabrics. Present research is based on analysis and comparison of the main concepts and conditions of textile technology and graph theory.

面向图嵌入的改进图注意机制模型

针对图神经网络模型学习图嵌入节点表示过程中易丢失大量特征信息及其图拓扑保留不完整的问题,提出了一种改进的图注意力(graph attention)机制模型。该模型分为节点级双向注意力机制和图级自注意图池化两部分。在学习图节点新的特征向量表示过程中,采取计算双向图注意力权重的方式,为邻域节点的保留提供可靠选择的同时增强节点间的相似属性;在图的整体拓扑上结合自注意图池,使用节点特征向量作为输入,通过注意卷积层提供的自注意权重在池化层生成图嵌入表示;在Cora、Citeseer、Pubmed数据集上进行了测试,实验结果表明:相比于基准图注意力机制模型,改进模型能够充分考虑图的局部和整体结构特征,有效增强模型聚合邻域信息的能力,减少了图嵌入过程中原始特征的丢失,明显提升了模型在下游任务的表现性能。

基于消息传递图神经网络的电力系统状态估计

近年来,随着电网规模日益增大,系统拓扑变化愈加频繁,拓扑变化组合向多样化趋势发展。然而,现有数据驱动状态估计模型仅能处理欧式数据,不能有效挖掘像拓扑信息这样的非欧式数据,因此现有数据驱动模型在拓扑频繁变化时的适应性较差。该文提出一种基于消息传递图神经网络(messagepassingneural network, MPNN)的电力系统状态估计模型。首先,利用拓扑参数和量测信息构建图数据集;其次,基于不同拓扑下的图数据训练消息传递图神经网络,得到状态估计模型;最后,在线应用时将该拓扑下的图数据输入已训练好的网络模型即可得到当前断面的状态量。通过对IEEE标准系统和中国某实际省网的算例测试,并将估计结果与加权最小二乘法、加权最小绝对值法以及深度神经网络算法和卷积神经网络算法进行比较。结果表明,该算法更能适应大规模电网中实时拓扑变化的特性。