Search

Automatic traffic accident detection has attracted the attention of the machine vision community for the rapid development of autonomous intelligent transportation systems (ITS). However, previous studies in this domain have been constrained by small-scale datasets with limited scope, impeding their effectiveness and applicability. Specifically, highway traffic accidents, often resulting in severe consequences due to higher speeds, require a more comprehensive approach to detection. The use of video surveillance provides a unique perspective, capturing the entire accident sequence. Unfortunately, existing traffic accident datasets are either not sourced from surveillance cameras, not publicly available, or not tailored for highway scenarios. An open-sourced traffic accident dataset with various scenes from surveillance cameras is in great need and of practical importance. To fulfill the above urgent need, we endeavor to collect abundant video data of real traffic accidents and propose a large-scale traffic accidents dataset, named TAD. Various experiments on image classification, video classification, and object detection tasks, using public mainstream vision algorithms or frameworks are conducted in this work to demonstrate the performance of different methods. The proposed dataset together with the experimental results are presented as a new benchmark to improve computer vision research, especially in ITS. The dataset is publicly available at https://github.com/UnicomAI/UnicomBenchmark/tree/main/TADBench.

Remote sensing semantic segmentation tasks aim to automatically extract land cover types by accurately classifying each pixel. However, large-scale hyperspectral remote sensing images possess rich spectral information, complex and diverse spatial distributions, significant scale variations, and a wide variety of land cover types with detailed features, which pose significant challenges for segmentation tasks. To overcome these challenges, this study introduces a U-shaped semantic segmentation network that combines global spectral attention and deformable Transformer for segmenting large-scale hyperspectral remote sensing images. First, convolution and global spectral attention are utilized to emphasize features with the richest spectral information, effectively extracting spectral characteristics. Second, deformable self-attention is employed to capture global-local information, addressing the complex scale and distribution of objects. Finally, deformable cross-attention is used to aggregate deep and shallow features, enabling comprehensive semantic information mining. Experiments conducted on a large-scale hyperspectral remote sensing dataset (WHU-OHS) demonstrate that: first, in different cities including Changchun, Shanghai, Guangzhou, and Karamay, DTSU-Net achieved the highest performance in terms of mIoU compared to the baseline methods, reaching 56.19%, 37.89%, 52.90%, and 63.54%, with an average improvement of 7.57% to 34.13%, respectively; second, module ablation experiments confirm the effectiveness of our proposed modules, and deformable Transformer significantly reduces training costs compared to conventional Transformers; third, our approach achieves the highest mIoU of 57.22% across the entire dataset, with a balanced trade-off between accuracy and parameter efficiency, demonstrating an improvement of 1.65% to 56.58% compared to the baseline methods.

The dimension of the operational-fixed-interval-scheduling-problem-modeled (OFISP-modeled) $\mathcal {NP}$ -hard satellite range scheduling problem (SRSP) extends rather than expands especially in large cases. This feature inspires us with the chain-splitting-solving-splicing (CSSS) approach to this OFISP-modeled SRSP under the curse of dimension. To boost the performance for scheduling the large-scale SRSP, we introduce the tricks to the algorithmic design from the inside out. The proposed method splits the original problem to small morsels with chain-splitting (CS) procedure to feed the route-reduction-based dynamic programming (R-DP) with the introduction of a novel scheduling element, the critical resource (CR), to expedite the execution for the optimal subsolution to the subproblem. At last, the standard DP (S-DP) splices the subsolutions into a complete optimal one. We encase the CR-based subproblem solving with the chain splitting and splicing framework. We show that it stunts the exponential explosion in time expenditure greatly with the CP by the mathematical analysis. We justify the efficiency of the proposed method with an application to a large-scale real-world SRSP instance. The proposed CSSS approach outperforms in comparison with several state-of-the-art algorithms. We obtained the optimal solution with the proposed method within reasonable time for the cases up to 3000 jobs. To the best of our knowledge, besides our research, other research into the large-scale SRSP is still absent. This status quo implies the potential of our research as a benchmark case for the would-be comparison with the other future research.

This paper studied a novel parallel machine rescheduling problem with controllable processing times under machine breakdown and precedence constraints. This problem is strongly $\mathcal {NP}$ -hard, and we modeled it as a mixed-integer problem (MIP). The primal problem is decomposed into the discrete subproblem and the continuous one. We treat the discrete with the dispatching rule and analyze the continuous in the terms of mathematical programming. Pros and cons from the analysis lead us to implement the commercial solver. The proposed method is capable of efficiency and nonzero initial state. We introduce transitive reduction to cull the redundant constraints out of its directed acyclic graph (DAG) representation. Transitive reduction extends the efficiency from the dispatching rule to the solver. The proposed method can do the predictive scheduling and pick up the partial solution left by the machine breakdown in the reactive session. As a result, the complete method solves the rescheduling problem with efficiency, and allows for large cases. At last, the computational results showed that this technique significantly brought down the time and RAM consumption in using the solver. This technique allows the scheduler to solve big instances with computational economy and efficiency.

The fusion of two or more traits in multimodal biometrics generally improves recognition accuracy. The question is, by how much? Large-scale databases are better suited for training deep learning models for better generalization and accuracy. Therefore, a large-scale multimodal database is beneficial. However, publicly available large-scale multimodal databases are scarce, especially for faces and voices. Again, because a face image is 2-D while a voice is 1-D, there is the challenge of the best way to fuse both. Therefore, improvements owing to fusion have hitherto yielded marginal improvements. This study proposes a semi-automated curation algorithm for the extraction of the faces and voices of target individuals in videos to create a large-scale face-voice database. The curation technique involves observing the positions at the time of the occurrence of the target subject’s faces and voices in videos. These positions are supplied to a MATLAB2017b script that detects the faces in the observed regions, crops, resizes, auto-labels, and writes them to the disk. A second MATLAB2017b script, extracts the audio content within the observed regions, auto-labels, and writes the voice segments to the disk. The created database named NaijaFaceVoice consists of 2,656 subjects with over 2 million faces and 195 hours of utterances. The database was employed to develop a large-scale recognition system that leveraged Convolutional Neural Networks. Robust fusion methods incorporating the proposed Spectrogram-Voting concept significantly improved performance achieving a record equal error rate of 0.0003519%, an improvement by a factor of over 450.

This study focuses on the simultaneous optimization of machines and buffers in split and merge production systems. The objective was to minimize the total investment cost under a minimum throughput rate and maximum cycle time constraints. It is challenging to solve this type of stochastic resource allocation problem due to the phenomenon of the combinatorial explosion search space and the inability to obtain closed-form expressions for the optimization model. In this paper, a decomposition-coordination method (DCM) is proposed to optimize the machine types used, the number of machines, and the capacities of buffers of general feed-forward topology systems efficiently and accurately. Instead of directly targeting large-scale systems, the DCM decomposes the original system into several small decoupled systems with added coordination variables and then separately optimizes each decomposed system. An optimal or near-optimal solution is obtained after several iterations of the decoupled system optimization and coordination variable updating. Moreover, we develop a simulated annealing algorithm and non-dominated sorting genetic algorithm-II as benchmark algorithms and provide a parameter calibration analysis of the two metaheuristics. Finally, comprehensive numerical experiments are performed to demonstrate the performances of the DCM, and a multifactorial experimental analysis is conducted to determine the influence of the split and merge system parameters on the performances of the DCM. The results confirmed that the scale of the system, complexity of topology, cycle time constraint, traffic intensity, price ratio, and their interactions significantly influenced the total cost obtained from the DCM, whereas the scale of the system, traffic intensity, and price ratio significantly affected the computation time.

The large-scale multi-objective optimization problem is characterized by a large decision space. How to design an efficient optimization algorithm that can search a large decision space and find the global optimum in the objective space is a very challenging problem at present. In order to solve this problem, this paper proposes a sampling strategy based on direction vectors, which takes into account both convergence and diversity. First, select some excellent individuals who are close to the ideal point based on the reference vector. Secondly, construct a three-way search direction vector using the boundary point and an additional center point, and execute a directional sampling strategy called the convergence-related sampling strategy to improve the convergence of the algorithm. After, the direction vector is constructed among excellent individuals and executes a directional sampling strategy called the diversity-related sampling strategy to maintain the diversity of the population. Finally, the adjustment strategy of the reference vector in the Reference Vector Guidance Algorithm (RVEA) is adopted to adjust the reference vector. Numerical experiments are performed on large-scale multi-objective benchmark problem sets with 500, 1000, and 2000 decision variables and compared with the state-of-the-art algorithms. Experimental results show that the algorithm proposed in this paper is effective and can obtain solutions that are significantly better than those of the compared algorithms.

A decoupling concept of near-field shrinking dielectric superstrate (NFSDS) is proposed for large-scale, wideband, and dual-polarized mm-wave arrays. An NFSDS with a thickness of 4 mm ( $0.32 \lambda _{0}$ at 24 GHz) is mounted seamlessly above the array, which shrinks the near field of the array elements to reduce the space wave coupling while slightly increasing the surface wave coupling of the arrays. By loading a superstrate with a certain thickness and low permittivity, the total coupling of the mm-wave arrays is reduced significantly. Periodic air holes are drilled through the superstrate to lower the NFSDS permittivity. An $8\times8$ mm-wave array is simulated, fabricated, and measured to verify the proposed decoupling concept. The simulated and measured coupling of the mm-wave array is reduced from −17 dB to lower than −23.2 dB from 24−29.5 GHz and lower than −25 dB in most of the band, respectively. The radiation patterns of the array before and after decoupling almost keep unchanged. Moreover, the NFSDS can efficiently improve the array beam scanning capability. The measured results align well with the simulated.

Word embedding is possessed by Natural language processing as a key procedure for semantically and syntactically manipulating the unlabeled text corpus. While this process represents the extracted features of corpus on vector space that enables to perform the NLP tasks such as summary generation, text simplification, next sentence prediction, etc. There exist some approaches for word embedding that consider co-occurrence and word frequency, such as Matrix Factorization, skip-gram, hierarchical-structure regularizer, and noise contrastive estimation. These approaches have created mature word vectors for most spoken languages in the world, on the other hand, the research community turned their minor attention towards the Urdu language having 231.3 million speakers. This paper focuses on creating Urdu word embedding. To perform this task, we used a dataset covering different categories of News such as Business, Sports, Health, Politics, Entertainment, Science, world, and others. This dataset was tokenized while creating 288 million tokens. Further, for word vector formation we utilized skip-gram also known as the word2vec model. The embedding was performed while limiting the vector dimensions to 100, 200, 300, 400, 500, 128, 256, and 512. For evaluation Wordsim-353 and Lexsim-999 annotated datasets were utilized. The proposed work achieved a 0.66 Spearman correlation coefficient value for wordsim-353 and 0.439 for Lexsim-999. The results were compared with state-of-the-art and were observed better.

In this article, we introduce a robust procedure for the extraction of passive rational macromodels of low-loss electromagnetic structures with massive port counts. Such structures pose inherent challenges that make standard macromodeling tools and approaches inadequate, mainly due to complexity and sensitivity at low frequency. The proposed approach involves a preprocessing stage in which port response data from a full-wave electromagnetic solver are regularized and extrapolated to dc using an asymptotic modal representation. The resulting data samples are then processed by a dedicated compression algorithm that represents the full set of input–output responses in terms of a few basis functions, which are constructed by enforcing an exact low-frequency modal asymptotic behavior, possibly including higher order dc zeros. These zeros are preserved in any stage of rational fitting and passivity enforcement, resulting in dc and low-frequency compliant compressed passive macromodels. Numerical results with up to 400 ports demonstrate the superior performance and accuracy of the computed models with respect to state-of-the-art approaches. In particular, the resulting models preserve their accuracy irrespective of the loading conditions, including the limit cases of short and open terminations. [ABSTRACT FROM AUTHOR]

RGBT tracking receives a surge of interest in the computer vision community, but this research field lacks a large-scale and high-diversity benchmark dataset, which is essential for both the training of deep RGBT trackers and the comprehensive evaluation of RGBT tracking methods. To this end, we present a La rge- ${s}$ cale ${H}$ igh-diversity $\text{b}{e}$ nchmark for short-term ${R}$ GBT tracking (LasHeR) in this work. LasHeR consists of 1224 visible and thermal infrared video pairs with more than 730K frame pairs in total. Each frame pair is spatially aligned and manually annotated with a bounding box, making the dataset well and densely annotated. LasHeR is highly diverse capturing from a broad range of object categories, camera viewpoints, scene complexities and environmental factors across seasons, weathers, day and night. We conduct a comprehensive performance evaluation of 12 RGBT tracking algorithms on the LasHeR dataset and present detailed analysis. In addition, we release the unaligned version of LasHeR to attract the research interest for alignment-free RGBT tracking, which is a more practical task in real-world applications. The datasets and evaluation protocols are available at: https://github.com/mmic-lcl/Datasets-and-benchmark-code. [ABSTRACT FROM AUTHOR]

Gaussian process classification (GPC) provides a flexible and powerful statistical framework describing joint distributions over function space. Conventional GPCs, however, suffer from: 1) poor scalability for big data due to the full kernel matrix and 2) intractable inference due to the non-Gaussian likelihoods. Hence, various scalable GPCs have been proposed through: 1) the sparse approximation built upon a small inducing set to reduce the time complexity and 2) the approximate inference to derive analytical evidence lower bound (ELBO). However, these scalable GPCs equipped with analytical ELBO are limited to specific likelihoods or additional assumptions. In this work, we present a unifying framework that accommodates scalable GPCs using various likelihoods. Analogous to GP regression (GPR), we introduce additive noises to augment the probability space for: 1) the GPCs with step, (multinomial) probit, and logit likelihoods via the internal variables and 2) particularly, the GPC using softmax likelihood via the noise variables themselves. This leads to unified scalable GPCs with analytical ELBO by using variational inference. Empirically, our GPCs showcase superiority on extensive binary/multiclass classification tasks with up to two million data points. [ABSTRACT FROM AUTHOR]

Generally, fingerprint-based indoor localization works inefficiently when deployed in a large-scale area. This is because it consumes massive resources and takes long processing time for searching the exact location in the large fingerprint database. Moreover, the changing environment can degrade overall performance. To tackle these problems, we propose an adaptive indoor localization system for a large-scale area. Our system consists of three main parts. First, our area classification algorithm is the key to overcome the problem caused by the large-scale area. It identifies an area of the user's queries whether they are outdoor or located in a specific building. Specifically, the algorithm can filter out the queries sent from outdoor or out-of-scope areas. Then, the information of this part is sent to the next part. Second, our fingerprint-based indoor localization algorithm can utilize the information from the first part by searching only the fingerprint in the specific building. This can significantly reduce searching space and processing time in order to localize the exact location. Third, our missing-BSSID detector algorithm detects the missing Basic Service Set Identifiers (BSSIDs) in the incoming query and updates a sampling database. This part is for our system to quickly adapt to the changing environment. We evaluated and deployed our system in a large-scale exhibition including 37 multi-floor buildings, covering 486,000 m2 and generating approximately 600,000 records of queries from users. In addition, we created a simulation to evaluate our system in the critically-changing environment. Our proposed system achieves high accuracy. More importantly, our area classification algorithm can significantly reduce the overall processing time compared to the previous work. Also, we showed that when applying our missing-BSSID detector algorithm to our system as well as other existing systems, the overall system performance can be significantly improved.

The design, testing and optimization of Vehicle to Everything (V2X), connected and automated driving and Intelligent Transportation Systems (ITS) and technologies requires mobility traces and traffic simulation scenarios that can faithfully characterize the vehicular mobility at the macroscopic and microscopic levels under large-scale and complex scenarios. The generation of accurate scenarios and synthetic traces requires a precise modelling approach, and the possibility to validate them against real-world measurements that are generally not available for large-scale scenarios. This limits the open availability of realistic and large-scale traffic simulation scenarios. The purpose of this paper is to present a large-scale and high-accuracy traffic simulation scenario. The scenario has been implemented over the open-source SUMO traffic simulator and is openly released to the community. The scenario accurately models the traffic flow, the traffic speed and the road’s occupancy for 9 full days of traffic over a 97 km freeway section. The scenario models mixed traffic with light and heavy vehicles. The simulation scenario has been calibrated using a unique dataset provided by the Spanish road authority and a novel learning-based and iterative traffic demand calibration technique for SUMO. This technique, referred to as Clone Feedback, is proposed for the first time in this paper and does not require a pre-calibration to generate realistic traffic demand. Clone Feedback can generate calibrated mixed traffic (light and heavy vehicles) using as input only traffic flow measurements. The results obtained show that Clone Feedback outperforms two reference techniques for calibrating the traffic demand in SUMO.

Among the flooding-based time synchronization protocols in wireless sensor networks, rapid-flooding outperforms slow-flooding in terms of synchronization speed. However, sometimes it is not realistic to apply rapid-flooding in dense or low-duty cycle networks due to its operational constraint that timing messages must be forwarded immediately after being received. On the contrary, slow-flooding protocols are free from this operational constraint, but they have a fundamental limitation that their convergence time increases with the size of the network. In this letter, we prove that this limitation can be overcome by sharing the compensated skew and timing information to the next hop nodes so that they can all be synchronized in the same round. The evaluation results demonstrate that we can achieve a comparable convergence speed of rapid-flooding protocols with the slow-flooding-based mechanism and can provide better accuracy for network-wide and per-hop synchronization. [ABSTRACT FROM AUTHOR]

Offspring generation plays an important role in evolutionary multiobjective optimization. However, generating promising candidate solutions effectively in high-dimensional spaces is particularly challenging. To address this issue, we propose an adaptive offspring generation method for large-scale multiobjective optimization. First, a preselection strategy is proposed to select a balanced parent population, and then these parent solutions are used to construct direction vectors in the decision spaces for reproducing promising offspring solutions. Specifically, two kinds of direction vectors are adaptively used to generate offspring solutions. The first kind takes advantage of the dominated solutions to generate offspring solutions toward the Pareto optimal set (PS) for convergence enhancement, while the other kind uses those nondominated solutions to spread the solutions over the PS for diversity maintenance. The proposed offspring generation method can be embedded in many existing multiobjective evolutionary algorithms (EAs) for large-scale multiobjective optimization. Experiments are conducted to reveal the mechanism of our proposed adaptive reproduction strategy and validate its effectiveness. Experimental results on some large-scale multiobjective optimization problems have demonstrated the competitive performance of our proposed algorithm in comparison with five state-of-the-art large-scale EAs. [ABSTRACT FROM AUTHOR]

Vehicle-to-Pedestrian (V2P) communication is essential to enable a reliable collision avoidance system for vehicles and vulnerable road users (VRUs). In order to develop a reliable communication system accurate channel models are required. However, no dedicated channel model for V2P communication in critical scenarios has been proposed so far. To develop a channel model, we conducted wideband channel sounding measurements considering a typical collision scenario between a vehicle and a pedestrian in line-of-sight (LoS) and non-LoS (NLoS) situations. The wideband measurement data is used to study the characteristics of large-scale or shadow fading, pathloss and diffraction loss. The spatial correlation of shadow fading is examined and different models are proposed. For NLoS scenario, a 3D scenario model is designed to simulate the diffraction at a row of parked vehicles. The diffraction loss due to parked vehicles is analyzed and modeled based on multiple knife-edge diffraction model combined with a two-ray pathloss model. The results show that the proposed diffraction model provides a good match to the measured pathloss.

Risk impacts on the operation of electric power distribution networks, when large amounts of distributed photovoltaics (PVs) are included, are gradually increasing to become a critical consideration for distribution networks' operations. In this paper, a risk assessment method is proposed which considers large-scale distributed PV feeding into distribution networks. Cluster partitioning is used to group large-scale distributed PV, to more accurately calculate the operational risk indicators for distribution networks with large-scale distributed PV compared to traditional risk assessment methods. A copula function is used to describe the correlation between the PV output and the load, in order to more accurately calculate power flow information compared to the traditional independent model of output and load. Node voltages over-limits and branch power flows outside boundaries are selected as the risk indicators, and a comprehensive risk indicator is introduced to more comprehensively and meaningfully evaluate the risk of distribution networks' operation. Also, the effects on the operational risk of the distribution network of load fluctuations, and, PV quantity, capacity, and location are analyzed and discussed. Finally, the IEEE34-bus system was simulated, and the results for different scenarios were calculated and analyzed. The simulation results demonstrate the effectiveness of the proposed method.

A relatively large body of work exists examining sex differences in expressiveness; however, there remains little research of differences in expressiveness associated with aging. Observational studies of facial expressivity across ages are limited in part due to the poor scalability of traditional research methods. We collected over 17,000 videos of natural facial behavior using the Internet and performed a large observational study of smiling responses of people ages 18 to 70 years. Using automated facial coding we quantified the presence of smiles as people watched a set of controlled mundane online content. The likelihood of smiles and the duration of smiles increased with age. We attribute this to greater expression of positive emotion in older people. Women smiled more than men over all and gender differences increased significantly with age. We question whether results may be influenced by the effect of age on the accuracy of the automated smile detection; however, validation on a large set of human coded videos shows that the observed effects were not due to smile detection performance. [ABSTRACT FROM AUTHOR]

The real-time simulation of large-scale subthalamic nucleus (STN)-external globus pallidus (GPe) network model is of great significance for the mechanism analysis and performance improvement of deep brain stimulation (DBS) for Parkinson’s states. This paper implements the real-time simulation of a large-scale STN-GPe network containing 512 single-compartment Hodgkin-Huxley type neurons on the Altera Stratix IV field programmable gate array (FPGA) hardware platform. At the single neuron level, some resource optimization schemes such as multiplier substitution, fixed-point operation, nonlinear function approximation and function recombination are adopted, which consists the foundation of the large-scale network realization. At the network level, the simulation scale of network is expanded using module reuse method at the cost of simulation time. The correlation coefficient between the neuron firing waveform of the FPGA platform and the MATLAB software simulation waveform is 0.9756. Under the same physiological time, the simulation speed of FPGA platform is 75 times faster than the Intel Core i7-8700K 3.70 GHz CPU 32GB RAM computer simulation speed. In addition, the established platform is used to analyze the effects of temporal pattern DBS on network firing activities. The proposed large-scale STN-GPe network meets the need of real time simulation, which would be rather helpful in designing closed-loop DBS improvement strategies. [ABSTRACT FROM AUTHOR]

Subset selection, aiming to select the best subset from a ground set with respect to some objective function, is a fundamental problem with applications in many areas, such as combinatorial optimization, machine learning, data mining, computer vision, information retrieval, etc. Along with the development of data collection and storage, the size of the ground set grows larger. Furthermore, in many subset selection applications, the objective function evaluation is subject to noise. We thus study the large-scale noisy subset selection problem in this paper. The recently proposed DPOSS algorithm based on multiobjective evolutionary optimization is a powerful distributed solver for large-scale subset selection. Its performance, however, has been only validated in the noise-free environment. In this paper, we first prove its approximation guarantee under two common noise models, i.e., multiplicative noise and additive noise, disclosing that the presence of noise degrades the performance of DPOSS largely. Next, we propose a new distributed multiobjective evolutionary algorithm called DPONSS for large-scale noisy subset selection. We prove that the approximation guarantee of DPONSS under noise is significantly better than that of DPOSS. We also conduct experiments on the application of sparse regression, where the objective evaluation is often estimated using a sample data, bringing noise. The results on various real-world data sets, whose size can reach millions, clearly show the excellent performance of DPONSS. [ABSTRACT FROM AUTHOR]

Albeit remarkable progress has been made to improve the accuracy and completeness of multi-view stereo (MVS), existing methods still suffer from either sparse reconstructions of low-textured surfaces or heavy computational burden. In this paper, we propose a Confidence-based Large-scale Dense Multi-view Stereo (CLD-MVS) method for high resolution imagery. Firstly, we formulate MVS as a multi-view depth estimation problem, and employ a normal-aware efficient PatchMatch stereo to estimate the initial depth and normal map for each reference view. A self-supervised deep learning method is then developed to predict the spatial confidence for multi-view depth maps, which is combined with cross-view consistency to generate the ground control points. Subsequently, a confidence-driven and boundary-aware interpolation scheme using static and dynamic guidance is adopted to synthesize dense depth and normal maps. Finally, a refinement procedure which leverages synthesized depth and normal as prior is conducted to estimate cross-view consistent surface. Experiments show that the proposed CLD-MVS method achieves high geometric completeness while preserving fine-scale details. In particular, it has ranked No. 1 on the ETH3D high-resolution MVS benchmark in terms of $F_{1}$ -score. [ABSTRACT FROM AUTHOR]

U-type assembly lines have become a mainstream mode in manufacturing because of the higher flexibility and productivity compared with straight lines. Since the balancing problem of a large-scale U-type assembly line is known to be NP-hard, effective mathematical model and evolutionary algorithm are needed to solve this problem. This paper reviews the research status of the related literature in recent years and presents a hybrid evolutionary algorithm, namely, modified ant colony optimization inspired by the process of simulated annealing, to reduce the possibility of being trapped in a local optimum for the balancing problem of stochastic large-scale U-type assembly line. A modified mathematical model for this balancing problem considering stochastic properties is formulated. Furthermore, comparisons with genetic algorithm and imperialist competitive algorithm are conducted to evaluate this proposed method. The results indicate that this proposed algorithm outperforms prior methods in this balancing problem.

Lattice reduction is a popular preprocessing strategy in multiple-input multiple-output (MIMO) detection. In a quest for developing a low-complexity reduction algorithm for large-scale problems, this paper investigates a new framework called sequential reduction (SR), which aims to reduce the lengths of all basis vectors. The performance upper bounds of the strongest reduction in SR are given when the lattice dimension is no larger than 4. The proposed new framework enables the implementation of a hash-based low-complexity lattice reduction algorithm, which becomes especially tempting when applied to large-scale MIMO detection. Simulation results show that, compared to other reduction algorithms, the hash-based SR algorithm exhibits the lowest complexity while maintaining comparable error performance. [ABSTRACT FROM AUTHOR]

Most existing person re-identification (re-id) methods are designed based on the artificial closed-set assumption that the probe and gallery identities are exactly overlapped with a small search pool. This leads to poor scalability in real-world applications where the task is often to re-id a small set of target people (i.e., watch-list) among a large search pool with unknown ID overlap, namely, an open-set deployment setting. In this paper, we firstly propose a new person re-id setting called Watch-List based Open-Set (WLOS) person re-id, which is characterised by the above open-set deployment and a watch-list available at the training stage. Then, we address such a under-studied WLOS problem by formulating a novel Task Dedicated Deep Hashing (TDDH) approach which learning a purpose-specific deep hash model particularly for the given target people in an efficient end-to-end manner. Extensive experiments on three large-scale re-id benchmarks are conducted to demonstrate the advantages and superiority of the TDDH over a wide range of the state-of-the-art hashing and re-id methods under the more realistic open-set setting. [ABSTRACT FROM AUTHOR]

To satisfy the robust and real-time requirements of power flow calculation for large-scale power systems, a globally convergent method is proposed with trust-region techniques, which shows satisfying robustness and convergence. Then, this method is combined with Newton's method to achieve a two-stage algorithm, benefiting from their different advantages. In the first stage, the proposed globally convergent method is used for searching power flow solution. When the values of state variables in a certain iteration are close enough to the real operational point, the algorithm enters the second stage to use Newton's method to achieve the solution. This two-stage algorithm can achieve an accurate solution for solvable cases, and can also achieve a least-square solution, which is an approximate solution for unsolvable cases. Numerical experiments show that the proposed globally convergent method and two-stage algorithm have better robustness and efficiency than the existing methods in previous research. They have universality for well- and ill-conditioned systems as well as the cases under ill-conditioned operational modes, heavy loads and inappropriate initial values. They can also handle different limit violations, such as reactive power limit in PV buses and voltage limit with tap changers action, which could significantly benefit real practice. [ABSTRACT FROM AUTHOR]

Real-time rendering of large-scale Web3D scenes was difficult to implement in virtual reality systems and geographic information system (GIS) in the past because of the technical constraints in CPU, memory, and network bandwidth. In this paper, a model management strategy was proposed based on triangular meshes, in which neighborhood buildings are considered as nodes and connected. Each node in the mesh has a set of level of detail (LOD) models, including high, medium, and low precision models. Besides a model file, the high precision LOD of the node can be a subtriangular mesh as well. The three-dimensional (3-D) models in a complex scene can be flexibly managed with some nested triangular meshes. The memory management strategy and the display management strategy were discussed in this paper. According to the experimental results, the proposed method effectively achieves the progressive downloading, dynamic loading, and real-time display for a large-scale 3-D scene. Its performance is better than the traditional methods. [ABSTRACT FROM AUTHOR]

A routing algorithm is the pivotal and core issue in network communication. Traditional networks often use distributed routing algorithms that lead to difficulties in controlling network routes and accessing information concerning global topology. A software-defined network (SDN) is a new network framework that separates the network control plane from the forwarding/data plane. SDNs can access global topology through SDN controllers and make forwarding decisions based on flow tables. In this paper, we propose a box-covering-based routing (BCR) algorithm using a renormalization method that is applied to research the fractal dimensions of large-scale networks. We divide the entire SDN network into some subnets using a box-covering method, find the shortest path in every subnet and between subnets, and embed the proposed BCR algorithm code inside the SDN controller. We then design a testbed based on the Ryu controller and the Mininet platform for this algorithm. The results show that the proposed BCR algorithm has good low-latency performance and is suitable for a large-scale SDN.

Reconstructing dense, volumetric models of real-world 3D scenes is important for many tasks, but capturing large scenes can take significant time, and the risk of transient changes to the scene goes up as the capture time increases. These are good reasons to want instead to capture several smaller sub-scenes that can be joined to make the whole scene. Achieving this has traditionally been difficult: joining sub-scenes that may never have been viewed from the same angle requires a high-quality camera relocaliser that can cope with novel poses, and tracking drift in each sub-scene can prevent them from being joined to make a consistent overall scene. Recent advances, however, have significantly improved our ability to capture medium-sized sub-scenes with little to no tracking drift: real-time globally consistent reconstruction systems can close loops and re-integrate the scene surface on the fly, whilst new visual-inertial odometry approaches can significantly reduce tracking drift during live reconstruction. Moreover, high-quality regression forest-based relocalisers have recently been made more practical by the introduction of a method to allow them to be trained and used online. In this paper, we leverage these advances to present what to our knowledge is the first system to allow multiple users to collaborate interactively to reconstruct dense, voxel-based models of whole buildings using only consumer-grade hardware, a task that has traditionally been both time-consuming and dependent on the availability of specialised hardware. Using our system, an entire house or lab can be reconstructed in under half an hour and at a far lower cost than was previously possible. [ABSTRACT FROM AUTHOR]

This letter considers optimal transmit beamforming for a sub-connected large-scale MISO system with RF chain and per-antenna power constraints. The system is configured such that each RF chain serves a group of antennas. For the hybrid scheme, necessary and sufficient conditions to design the optimal digital and analog precoders are provided. It is shown that, in the optimum, the optimal phase shift at each antenna has to match the channel coefficient and the phase of the digital precoder. In addition, an iterative algorithm is provided to find the optimal power allocation. We study the case where the power constraint on each RF chain is smaller than the sum of the corresponding per-antenna power constraints. Then, the optimal power is allocated based on two properties: each RF chain uses full power and if the optimal power allocation of the unconstraint problem violates a per-antenna power constraint then it is optimal to allocate the maximal power for that antenna. [ABSTRACT FROM AUTHOR]

This paper tackles a large-scale and uncertain scheduling problem of semiconductor wafer fabrication system. Two kinds of uncertainties are considered here, including the machine breakdowns and the fluctuations of processing times. Instead of using the traditional rigid schedule, we propose a novel concept of operation-group-based soft schedule which is used as the new decision variable of the studied problem. The inherent idea of this concept is to make some critical scheduling decisions at the beginning of the scheduling horizon, and allow the remaining decisions to be made during the execution of the initial soft schedule. Furthermore, an operation-group-based soft scheduling approach (OGSSA) is proposed to deal with the uncertainties, which contains an offline optimization layer and an online dispatching layer. In the offline optimization layer, a prediction-based decomposition scheduling method is proposed to generate a soft schedule, and a global scheduling objective prediction model is constructed to evaluate the soft schedule. Then, the initial soft schedule is released to the online dispatching layer, and an online heuristic rule is designed to decide in real time, which operations and when to process. The computational results on the practical production data demonstrate the effectiveness of OGSSA under uncertain production environments. [ABSTRACT FROM AUTHOR]

Retrieving videos from large repositories using image queries is important for many applications, such as brand monitoring or content linking. We introduce a new retrieval architecture, in which the image query can be compared directly with database videos—significantly improving retrieval scalability compared with a baseline system that searches the database on a video frame level. Matching an image to a video is an inherently asymmetric problem. We propose an asymmetric comparison technique for Fisher vectors and systematically explore query or database items with varying amounts of clutter, showing the benefits of the proposed technique. We then propose novel video descriptors that can be compared directly with image descriptors. We start by constructing Fisher vectors for video segments, by exploring different aggregation techniques. For a database of lecture videos, such methods obtain a two orders of magnitude compression gain with respect to a frame-based scheme, with no loss in retrieval accuracy. Then, we consider the design of video descriptors, which combine Fisher embedding with hashing techniques, in a flexible framework based on Bloom filters. Large-scale experiments using three datasets show that this technique enables faster and more memory-efficient retrieval, compared with a frame-based method, with similar accuracy. The proposed techniques are further compared against pre-trained convolutional neural network features, outperforming them on three datasets by a substantial margin. [ABSTRACT FROM AUTHOR]

To prolong the lifetime of energy constrained devices in Internet of Things, devices can harvest wireless energy from the control signal multicast from the access point. Unfortunately, hampered by the path-loss, the efficiency of such multicast wirelessly powered network is low. While large-scale antennas at access point can be used to improve the efficiency, the beamforming design problem in multicast wirelessly powered network is known to be NP-hard, and the traditional difference of convex programming becomes prohibitively time consuming in large-scale settings. On the other extreme, by using the assumption of infinite number of antennas and applying the law of large numbers, simple beamforming solution is possible. However, when applied to scenarios with finite number of antennas, the performance of such asymptotic solution is far from that of difference of convex programming. To resolve this apparent complexity-performance dilemma, this paper develops an algorithm which reduces the computation time by orders of magnitude, while still guaranteeing the same performance compared with the difference of convex programming. In particular, the proposed algorithm consists of two fast-convergent iterative procedures and is guaranteed to obtain a Karush–Kuhn–Tucker solution. Furthermore, in each iteration, the algorithm only requires the computation of inner products between channel vectors and can be run in parallel for all the users. Thus, the complexity scales linearly with the number of antennas at access point. Finally, numerical results validate the performance and the speed of the proposed scheme. [ABSTRACT FROM AUTHOR]

This paper proposes three simple, compact yet effective representations of depth sequences, referred to respectively as dynamic depth images (DDI), dynamic depth normal images (DDNI), and dynamic depth motion normal images (DDMNI), for both isolated and continuous action recognition. These dynamic images are constructed from a segmented sequence of depth maps using hierarchical bidirectional rank pooling to effectively capture the spatial-temporal information. Specifically, DDI exploits the dynamics of postures over time, and DDNI and DDMNI exploit the 3-D structural information captured by depth maps. Upon the proposed representations, a convolutional neural network (ConvNet)-based method is developed for action recognition. The image-based representations enable us to fine-tune the existing ConvNet models trained on image data without training a large number of parameters from scratch. The proposed method achieved the state-of-art results on three large datasets, namely, the large-scale continuous gesture recognition dataset (means the Jaccard index 0.4109), the large-scale isolated gesture recognition dataset ( \text59.21\%), and the NTU RGB+D dataset ( \text87.08\% cross-subject and \text84.22\% cross-view) even though only the depth modality was used. [ABSTRACT FROM PUBLISHER]

In this paper, we propose a framework for privacy-preserving outsourced functional computation across large-scale multiple encrypted domains, which we refer to as POFD. With POFD, a user can obtain the output of a function computed over encrypted data from multiple domains while protecting the privacy of the function itself, its input and its output. Specifically, we introduce two notions of POFD, the basic POFD and its enhanced version, in order to tradeoff the levels of privacy protection and performance. We present three protocols, named Multi-domain Secure Multiplication protocol (MSM), Secure Exponent Calculation protocol with private Base (SECB), and Secure Exponent Calculation protocol ( SEC), as the core sub-protocols for POFD to securely compute the outsourced function. Detailed security analysis shows that the proposed POFD achieves the goal of calculating a user-defined function across different encrypted domains without privacy leakage to unauthorized parties. Our performance evaluations using simulations demonstrate the utility and the efficiency of POFD. [ABSTRACT FROM AUTHOR]

Fast and accurately categorizing the millions of aerial images on Google Maps is a useful technique in pattern recognition. Existing methods cannot handle this task successfully due to two reasons: 1) the aerial images’ topologies are the key feature to distinguish their categories, but they cannot be effectively encoded by a conventional visual codebook and 2) it is challenging to build a realtime image categorization system, as some geo-aware Apps update over 20 aerial images per second. To solve these problems, we propose an efficient aerial image categorization algorithm. It focuses on learning a discriminative topological codebook of aerial images under a multitask learning framework. The pipeline can be summarized as follows. We first construct a region adjacency graph (RAG) that describes the topology of each aerial image. Naturally, aerial image categorization can be formulated as RAG-to-RAG matching. According to graph theory, RAG-to-RAG matching is conducted by enumeratively comparing all their respective graphlets (i.e., small subgraphs). To alleviate the high time consumption, we propose to learn a codebook containing topologies jointly discriminative to multiple categories. The learned topological codebook guides the extraction of the discriminative graphlets. Finally, these graphlets are integrated into an AdaBoost model for predicting aerial image categories. Experimental results show that our approach is competitive to several existing recognition models. Furthermore, over 24 aerial images are processed per second, demonstrating that our approach is ready for real-world applications. [ABSTRACT FROM PUBLISHER]

Implantable wireless neural recording from a large ensemble of simultaneously acting neurons is a critical component to thoroughly investigate neural interactions and brain dynamics from freely moving animals. Recent researches have shown the feasibility of simultaneously recording from hundreds of neurons and suggested that the ability of recording a larger number of neurons results in better signal quality. This massive recording inevitably demands a large amount of data transfer. For example, recording 2000 neurons while keeping the signal fidelity ($>12$ bit, $> 40$ KS/s per neuron) needs approximately a 1-Gb/s data link. Designing a wireless data telemetry system to support such (or higher) data rate while aiming to lower the power consumption of an implantable device imposes a grand challenge on neuroscience community. In this paper, we present a wireless gigabit data telemetry for future large-scale neural recording interface. This telemetry comprises of a pair of low-power gigabit transmitter and receiver operating at 60 GHz, and establishes a short-distance wireless link to transfer the massive amount of neural signals outward from the implanted device. The transmission distance of the received neural signal can be further extended by an externally rendezvous wireless transceiver, which is less power/heat-constraint since it is not at the immediate proximity of the cortex and its radiated signal is not seriously attenuated by the lossy tissue. The gigabit data link has been demonstrated to achieve a high data rate of 6 Gb/s with a bit-error-rate of 10−12 at a transmission distance of 6 mm, an applicable separation between transmitter and receiver. This high data rate is able to support thousands of recording channels while ensuring a low energy cost per bit of 2.08 pJ/b. [ABSTRACT FROM PUBLISHER]