一种基于二分图最优匹配的镜头检索方法

镜头检索是基于内容的视频检索的重要内容.本文首次尝试将二分图的最优匹配用于镜头检索.与现有方法相比,本文提出的方法强调在一一对应的前提下,全面客观地度量两个镜头的相似度.把两个镜头的相似度度量建模为一个带权的二分图:镜头中的每一帧看成二分图的一个结点,两个镜头之间任意帧的相似值作为边的权值.在一一对应的前提下,利用最优匹配的Kuhn-Munkres算法求出该二分图的最大权,以此作为两个镜头的相似度.考虑到检索速度问题,提出了两个改进算法.实验对比结果证实了本文所提方法在镜头检索中的优异表现.     

异构网络融合环境下基于加权二分图的网络选择算法

摘要：在异构网络融合场景中,为了合理利用异构网络资源,提出了一种基于加权二分图的网络选择算法。将网络选择过程映射为寻找最优加权二分图的过程,用户和网络双方的需求经数学建模后进行供需最优匹配,供需匹配有二次选择机会以平衡网络负载。与其他算法相比,提出的算法可以为不同QoS业务提供满足需求的匹配结果,并且能够动态调整网络负载,实现异构网络融合场景中用户需求和网络资源的合理匹配。     

Robust multi-view correspondence of noisy 2D points usingrelaxation

An efficient ghost removal algorithm for reliably extracting correspondence information for 2D points from several views using epipolar constraints is proposed. The point matching relations are modelled using a weighted k-partite graph and a globally convergent iterative algorithm is used to extract reliable maximum cliques from the graph using probabilistic relaxation labelling. Experimental results show the effectiveness of the proposed method     

基于卷积神经网络的高效三维模型检索方法

目前基于视图的三维模型检索已经成为一个研究热点.该方法首先将三维模型表示为二维视图的集合,然后采用深度学习技术进行分类和检索.但是现有的方法在精度和效率方面都有待提升.本文提出了一种新的三维模型检索方法,该方法包括索引建立和模型检索.在索引建立阶段,选择代表性视图输入到训练好的卷积神经网络(Convolutional Neural Network,CNN)中以提取特征和分类,进而根据特征类别对特征进行组织以建立索引在检索阶段,使用CNN和投票算法将输入模型的代表性视图分类为一个类别,然后仅选择这个类别的特征而不是所有类别的特征进行相似度匹配,因此减少了搜索空间.此外,随着用于检索的视图数量的逐渐增加,一旦可以确定三维模型,检索过程将提前终止.实验的数据选用刚性三维模型数据集ModelNet10,ModelNet40和非刚性三维模型数据集McGill10.结果表明,该方法在提升检索效率的同时,确保检索准确率分别高达94％、92％和100％.     

改进GCNs在指静脉特征表达中的应用

基于图模型的指静脉全局特征表达方法不仅可以降低成像质量对采集设备的依赖性,还能提高匹配效率。针对于目前指静脉图模型的研究中存在的图结构不稳定,匹配效率随图模型的变大而降低的问题,本文提出了一种基于SLIC(Simple Linear Iterative Clustering)超像素分割算法构建加权图的方法,并改进ChebyNet图卷积神经网络(Graph Convolutional Neural Networks, GCNs)提取加权图的图级(graph-level)特征。针对指静脉样本数普遍较少,而ChebyNet中卷积网络参数量较大容易造成过拟合以及其快速池化层不能自适应地选择节点的问题,本文提出了全局池化结构的改进GCNs模型SCheby-MgPool(Simplified Cheby-Multi gPool)。实验结果表明,本文提出的方法提取的指静脉特征在识别精度,匹配效率上都具有较好的性能。      

Correspondence Propagation with Weak Priors

For the problem of image registration, the top few reliable correspondences are often relatively easy to obtain, while the overall matching accuracy may fall drastically as the desired correspondence number increases. In this paper, we present an efficient feature matching algorithm to employ sparse reliable correspondence priors for piloting the feature matching process. First, the feature geometric relationship within individual image is encoded as a spatial graph, and the pairwise feature similarity is expressed as a bipartite similarity graph between two feature sets; then the geometric neighborhood of the pairwise assignment is represented by a categorical product graph, along which the reliable correspondences are propagated; and finally a closed-form solution for feature matching is deduced by ensuring the feature geometric coherency as well as pairwise feature agreements. Furthermore, our algorithm is naturally applicable for incorporating manual correspondence priors for semi-supervised feature matching. Extensive experiments on both toy examples and real-world applications demonstrate the superiority of our algorithm over the state-of-the-art feature matching techniques.      

Indexation de modèles 3D par graphe de Reeb multirésolution augmenté

This article presents an indexing method for 3D mesh models applied to shape and appearance similarity retrieval in database of 3D objects. The approach relies on the multi-resolution Reeb graph matching method proposed by Hilaga et al. in 2001 [1]. In our framework, we consider 3D mesh models of various geometrical complexity, different resolution, and sometimes textured. The original approach, based on the 3D object topology, is not accurate enough to obtain satisfying matching. Therefore we propose to reinforce the topological consistency conditions of the matching and to merge within the graph geometrical and visual information to improve the matching and the calculation of the similarity between models. Besides, all these new attributes can be freely weighted to fit the user requirements for object retrieval. We obtain a flexible multiresolutional and multicriteria representation that we called augmented Multiresolution Reeb Graph (aMrg). The approach has been tested and compared with other methods. It reveals very performant for the retrieval and the classification of similar 3D shapes.     

Statistical modeling and many-to-many matching for view-based 3D object retrieval

We address the task of view-based 3D object retrieval, in which each object is represented by a set of views taken from different positions, rather than a geometrical model based on polygonal meshes. As the number of views and the view point setting cannot always be the same for different objects, the retrieval task is more challenging and the existing methods for 3D model retrieval are infeasible. In this paper, the information in the sets of views is exploited from two aspects. On the one hand, the form of histogram is converted from vector to state sequence, and Markov chain (MC) is utilized for modeling the statistical characteristics of all the views representing the same object. On the other hand, the earth mover's distance (EMD) is involved to achieve many-to-many matching between two sets of views. For 3D object retrieval, by combining the above two aspects together, a new distance measure is defined, and a novel approach to automatically determine the edge weights in graph-based semi-supervised learning is proposed. Experimental results on different databases demonstrate the effectiveness of our proposal.     

Objective image quality assessment of 3D synthesized views

Depth-Image-Based-Rendering (DIBR) techniques are essential for three-dimensional (3D) video applications such as 3D Television (3DTV) and Free-Viewpoint Video. However, this process is based on 3D warping and can induce serious distortions whose impact on the perceived quality is far different from the one experienced in the 2D imaging processes. Since quality evaluation of DIBR-synthesized views is fundamental for the design of perceptually friendly 3D video systems, an appropriate objective quality metric targeting the assessment of DIBR-synthesized views is momentous. Most of the 2D objective quality metrics fail in assessing the visual quality of DIBR-synthesized views because they have not been conceived for addressing the specificities of DIBR-related distortions. In this paper, a new full-reference objective quality metric, 3DSwIM (3D Synthesized view Image Quality Metric), dedicated to artifacts detection in DIBR-synthesized view-points is presented. The proposed scheme relies on a comparison of statistical features of wavelet subbands of two input images: the original image and the DIBR-based synthesized image. A registration step is included before the comparison step so that best matching blocks are always compared to ensure “shifting-resilience”. In addition, a skin detection step weights the final quality score in order to penalize distorted blocks containing “skin-pixels” based on the assumption that a human observer is most sensitive to impairments affecting human subjects. Experimental tests show that the proposed method outperforms the conventional 2D and DIBR-dedicated quality metrics under test.     

Research on Base Station Sleeping Mechanism of User Connections in Ultra Dense Network

On the basic of traditional macro cellular networks, ultra dense networks deploy plenty of low-power nodes working with maximum power, which provide superior communication quality and produce more energy consumption in networks. Aiming at the problem of wasting network resources caused by low-power nodes during low-load period in ultra dense networks, we study a kind of base station sleeping mechanism based on user connections. When the network loads are low, the connections between users and base stations (BSs) are sensed by environment awareness technology. Then, the connection relationship is used to establish a connection matrix. After that we use the established connection matrix to build a weighted bipartite graph. Taking the users’ QoS and the load of BSs into account, we build a weight matrix by weighting the two as the weights of the bipartite graph. We get the optimal connection between users and base stations by constantly optimizing the value of the connection matrix. By modeling, we transform the optimization problem into a 0–1 integer programming problem and get the optimal connection by Particle Swarm Optimization algorithm. Finally, the sleeping mechanism is executed according to this connection matrix, and we achieve the goal of energy conservation by closing low-load BSs.     

VLSI中双层布线导孔数的优化算法

在超大规模集成电路的设计中,通常需要尽可能地减少布线时产生的导孔,因为导孔数是影响集成电路成品率、可靠性及各种电性能的重要因素.本文的工作是解决VLSI双层金属布线的导孔数优化问题.实际的布线根据其拓扑特性和电连通性等效为带权的图.将这种方法推广到三联和四联导孔的情形,并引入了添加冗余导孔的方法以进一步减少导孔的数目.最后给出了对变权图求最大偶子图的算法,以便对三联及四联情形的导孔进行优化.算法实现后试算的实例表明一般能减少30％-50％的导孔.     

Genetic algorithm for model-based matching of projected images of three-dimensional objects

A novel genetic algorithm (GA) is proposed for searching for the existence of a projective transform which, when applied to the model, results in the best alignment with an unknown 2D edge image. The presence of a valid solution reflects that the latter can be regarded as one of the projected views of the model. On this basis, the identity of an unknown edge image can be deduced by matching it against a set of 3D reference models. To increase the efficiency of the process, a two-pass, coarse-to-fine strategy is adopted. Initially, an unknown image is first classified to a small group of models by matching their outermost boundaries. Next, a fine but slower matching algorithm selects model(s) that share similar internal edge features as the unknown image. In the design of the method, the authors have adopted an alternative projective transform representation that involves fewer parameters and allows constraints to be easily imposed on their dynamic ranges. This effectively lowers the chance of premature convergence and increases the success rate. Experimental results obtained with the proposed scheme are encouraging and demonstrate the feasibility of the approach.     

Data dissemination scheduling algorithm for V2R/V2V in multi-channel VANET

Considering that the data dissemination in multi-channel VANET (vehicular ad hoc network),a cooperative data dissemination scheduling algorithm was introduced for V2R(vehicle to roadside unit) and V2V(vehicle to vehicle).The algorithm created initial scheduling operators according to data requisition of vehicles.Then,initial collision graph and collision matrix were generated based on the conflict among initial scheduling operators.After proving the positive semidefinite of collision matrix,SDP (semidefinite programming) was used to channel allocation and collision graph creation.The algorithm then assigned weights for each data requisition according to dwell time and data volume of vehicles in RSU service region.Furthermore,it selected maximum weighted independent set of collision graph.The goal was to satisfy the most urgent data requisitions by V2R/V2V cooperate transmission.Transportation simulation results demonstrate that the proposed solution effectively promotes the service capacity by utilizes the multichannel of VANET and V2R/V2V transmission scheduling.     

Yield enhancement techniques for 3-dimensional random access memories

As the advances of process technology keep growing, three-dimensional (3D) integration with through silicon vias is a new alternative solution to extend Moore’s law especially for random access memories (RAMs). In general, the reliability and fabrication yield of the traditional 2D memories can be improved by the incorporation of some form of redundancy. However, for 3D integration, the scenarios for the repair process are totally different. The redundancy exclusively added in a memory tier can also be reused to repair defects in the other memory tier after the bonding process. That is, the concept of inter-tier redundancy can be exploited to further increase the yield of 3D memories. Die-to-die and die-to-wafer bonding can be adopted. In this paper, we propose an efficient die-stacking flow and the corresponding built-in self-repair architectures for yield enhancement of 3D memories. The matching problem for die stacking can be converted into a bipartite graph maximal matching problem and the traditional algorithm can be used to solve this problem. Experimental results show that the proposed stacking flow, algorithm, and the corresponding BISR (built-in self-repair) architecture can improve fabrication yield significantly.     

Camera constraint-free view-based 3-D object retrieval

Recently, extensive research efforts have been dedicated to view-based methods for 3-D object retrieval due to the highly discriminative property of multiviews for 3-D object representation. However, most of state-of-the-art approaches highly depend on their own camera array settings for capturing views of 3-D objects. In order to move toward a general framework for 3-D object retrieval without the limitation of camera array restriction, a camera constraint-free view-based (CCFV) 3-D object retrieval algorithm is proposed in this paper. In this framework, each object is represented by a free set of views, which means that these views can be captured from any direction without camera constraint. For each query object, we first cluster all query views to generate the view clusters, which are then used to build the query models. For a more accurate 3-D object comparison, a positive matching model and a negative matching model are individually trained using positive and negative matched samples, respectively. The CCFV model is generated on the basis of the query Gaussian models by combining the positive matching model and the negative matching model. The CCFV removes the constraint of static camera array settings for view capturing and can be applied to any view-based 3-D object database. We conduct experiments on the National Taiwan University 3-D model database and the ETH 3-D object database. Experimental results show that the proposed scheme can achieve better performance than state-of-the-art methods.     

Relax online resource allocation algorithms for D2D communication

Maximizing the system sumrate by sharing the resource blocks among the cellular user equipments and the D2D (device to device) pairs while maintaining the quality of service is an important research question in a D2D communication underlaying cellular networks. The problem can be optimally solved in offline by using the weighted bipartite matching algorithm. However, in long‐term evolution and beyond (4G and 5G) systems, scheduling algorithms should be very efficient where the optimal algorithm is quite complex to implement. Hence, a low complexity algorithm that returns almost the optimal solution can be an alternative to this research problem. In this paper, we propose 2 less complex stable matching–based relax online algorithms those exhibit very close to the optimal solution. Our proposed algorithms deal with fixed number of cellular user equipments and a variable number of D2D pairs those arrive in the system online. Unlike online matching algorithms, we consider that an assignment can be revoked if it improves the objective function (total system sumrate). However, we want to minimize the number of revocation (ie, the number of changes in the assignments) as a large number of changes can be expensive for the networks too. We consider various offline algorithms proposed for the same research problem as relaxed online algorithms. Through extensive simulations, we find that our proposed algorithms outperform all of the algorithms in terms of the number of changes in assignment between 2 successive allocations while maintaining the total system sumrate very close to the optimal algorithm.     

改进型形状上下文的工件立体匹配方法

为了解决传统形状上下文直方图相似性计算需遍历轮廓采样点、计算复杂性高、不能满足工业实时性要求的问题,提出了一种改进型形状上下文的工件立体匹配方法。改进匹配搜索策略,引入立体图像对质心的视差约束条件,并利用形状上下文直方图分布信息进行初步筛选,获取候选匹配点集,减少后续匹配计算复杂度;为增加匹配点与非匹配点的区分度,对形状上下文相似性度量进行加权处理;匹配时融合对应轮廓采样点33邻域的梯度方向直方图特征,与形状上下文构成联合相似性度量;最后采用随机抽样一致性算法剔除误匹配点对。对改进算法进行了理论分析,并应用于工件图像进行实验验证,通过实验给出了相应的实验对比。结果表明,改进的方法具有较高的匹配精度和更快的匹配速度,鲁棒性高。     

基于两步加权L1范数约束的高分辨率波达方向和功率估计

该文提出一种基于两步加权L1范数约束的波达方向(DOA)和功率联合估计新算法。该算法首先对阵列协方差矩阵元素进行求和平均运算获得统计性能更优的向量观测模型,进而在相应的过完备基下获得观测模型的稀疏表示。利用MUSIC谱函数和初始估计结果作为权值构建两步加权L1范数约束稀疏重构方法,重构获得多信源的DOA和功率估计。通过2分布函数选择了合理的正则化参数。计算机仿真结果显示,与同类其它方法相比,该文所提算法具有更高的分辨率和估计精度。     

Multi-stage online task assignment driven by offline data under spatio-temporal crowdsourcing

In the era of the Internet of Things (IoT), the crowdsourcing process is driven by data collected by devices that interact with each other and with the physical world. As a part of the IoT ecosystem, task assignment has become an important goal of the research community. Existing task assignment algorithms can be categorized as offline (performs better with datasets but struggles to achieve good real-life results) or online (works well with real-life input but is difficult to optimize regarding in-depth assignments). This paper proposes a Cross-regional Online Task (CROT) assignment problem based on the online assignment model. Given the CROT problem, an Online Task Assignment across Regions based on Prediction (OTARP) algorithm is proposed. OTARP is a two-stage graphics-driven bilateral assignment strategy that uses edge cloud and graph embedding to complete task assignments. The first stage uses historical data to make offline predictions, with a graph-driven method for offline bipartite graph matching. The second stage uses a bipartite graph to complete the online task assignment process. This paper proposes accelerating the task assignment process through multiple assignment rounds and optimizing the process by combining offline guidance and online assignment strategies. To encourage crowd workers to complete crowd tasks across regions, an incentive strategy is designed to encourage crowd workers’ movement. To avoid the idle problem in the process of crowd worker movement, a drop-by-rider problem is used to help crowd workers accept more crowd tasks, optimize the number of assignments, and increase utility. Finally, through comparison experiments on real datasets, the performance of the proposed algorithm on crowd worker utility value and the matching number is evaluated.