Search

Recently, collaborative representation based classifiers (CRC) have shown outstanding performances in recognition tasks. The key to success of most CRC algorithms states that the testing samples can be coded well by a suitable dictionary globally, while the local knowledge between samples has not been fully considered. We observe that the representations of similar samples have a high degree of similarity. In order to take advantage of this important similarity information, this paper proposes a new local knowledge-based collaborative representation model for image classification. Specifically, certain adjacent training samples of the testing image should be determined firstly, and then the representations of these neighborhoods can be applied to guide the coefficients of the testing samples to be more discriminative. Further, we derive a robust version of the proposed method to treat the face recognition with occlusions or corruptions. Extensive experiments are carried out to show the superiority of the proposed method over other state-of-the-art classifiers on various image recognition tasks.

Micro-flow meters are expanding their flow range and improving measurement accuracy, which brings new challenges to their calibration. For micro-flow calibration, one of the challenges is to maintain a stable and accurate flow through the calibration runs. Other challenges also include evaporation, degassing, surface tension, and temperature variation. This paper combines the theories with experiments to analyze the source of each uncertainty component and deal with them in the control system. Most importantly, a new diverter valve has been developed to specifically address the challenge. This novel design can accomplish both the standing-start-and-finish method and the flying-start-and-finish method. As a result, the system can calibrate flow in the range of 0.1-80g/min, and the expanded uncertainty of the calibration rig was estimated to be 0.12% at 95% confidence level.

As with the most of established magnetic inductive flowmeters(MIF) methods for measuring the level of the measuring line, the accuracy of induction flowmeters are strongly affected by the shape of the electrodes. In order to improve the sensitivity of measurement of the filling degree with the flow in the partially full conduit, the position and the circumference of measuring line that the electrodes encompass should be taken into consideration. In this paper, the electrodes are optimized by improving the sensitivity of the measurement of the fluid level and manipulating the weight function, which reduces the influence by the distorted flow profiles. A new finite element procedure is presented to study the voltage output between various shapes of the horizontally-opposed electrodes. Then, the relationship between the sensor sensitivity and fluid level is further obtained. Optimal shape of the electrodes is obtained based on the sensitivity and weight function theory.

With the development of technology, especially the rise of some practical application technology such as AR, VR, and human-computer interaction, 3D human pose estimation has shown more and more extensive prospects. However, in existing 3D human pose estimation methods, there are still some issues. For example, the lack of multi-scale and diversity in the analysis of human posture leads to insufficient refinement in fitting and segmentation of body parts, and eventually causing unsatisfactory performance. In this paper, we propose a feature-fusion-based dense estimation model (FFD model) for 3D human pose estimation to solve the above problems. Specifically, we design a feature fusion module obtaining multi-scale feature information, which not only captures the overall information about the human posture, but also retains the position information and detailed information of the human body part. Furthermore, we use a dense feature extraction backbone network to expand the diversity of receptive fields, thus making the human pose estimation more refined and precise. Experiments on the COCO dataset show the superiority of FFD model compared with the baseline model.

The multiple-preys pursuit (MPP) is the adversarial game between predators and preys. If the capture of a prey is defined as that it cannot move anymore due to the surrounding of predators, there are two kinds of task allocations. One is about assigning which prey to which group of predators so that all preys can be captured. The other is about assigning which capturing position to which predator to encircle the prey simultaneously. In this paper, the MPP is modeled as a dynamic optimization problem and each its time step is solved in two stages. Firstly, the first kind of task allocation problem is modeled as the biquadratic assignment problem (BiQAP) and a MPP fitness function is proposed for the evaluation of such BiQAP task allocations. In this way, the MPP is transformed to several single-prey pursuit (SPP) problems. Secondly, for each SPP, we extend the coordinated SPP strategy CCPSO-R (cooperative coevolutionary particle swarm optimization for robots) to its parallel version as PCCPSO-R to enable the parallel implicit capturing position allocating by parallel observation, decision making, and moving of predators. Through experiments of the current BiQAP solvers on the task allocation, we improve the best one of them in statistic based on the domain knowledge. Moreover, the advantages of PCCPSO-R in the capturing efficiency over CCPSO-R is testified in the MPP experiments.

This paper analyzes the convergence of iterative learning control for a class of discrete-time systems over networks with successive input data compensation. Specifically, the successively dropped input data in current iterations are compensated by the one actuator received correctly with the same time instant label in the latest iteration. Through analyzing the variation of elements in the transition matrices of the input errors at the controller side, the convergence of output errors is addressed. The analysis shows the selection range of learning gain is determined by the maximum number of input data dropped successively. Moreover, the convergence of system with successive input data compensation is guaranteed by trading the convergence speed, and the more input data are successively compensated in iteration domain, the more convergence speed of the system is reduced. Finally, numerical experiments are given to corroborate the theoretical analysis.

AT power supply mode is widely used in high-speed railway. This paper introduces three kinds of AT power supply modes, and combined with the current situation of the harmonic problems in traction network caused by the mixed running of AC-DC electric locomotive and AC-DC-AC EMU, carries out the simulation comparative study on the harmonic characteristics of traction network of multi-vehicle mixed running lines under three kinds of AT power supply modes. Taking SS6B locomotive and CRH2 EMU, which are widely used in the actual situation at present, for examples, the interactive simulation model between the vehicle and traction network is established by using Matlab/Simulink simulation software. Through the simulation analysis of the harmonic characteristics of SS6B electric locomotive and CRH2 EMU, the correctness of the comparison conclusion in this paper is verified. On the basis of the vehicle network interactive simulation model, the harmonic characteristics of the traction network under different traction network lengths, different locomotive positions and different locomotive working conditions of three AT power supply modes were compared and analyzed. The study results show that the vehicle network interactive simulation model established by the three AT power supply modes can accurately reflect the influence of AC-DC locomotive mixed running on the harmonic characteristics of traction network.

With the continuous popularization of smart devices, a new computing paradigm–Spatial Crowdsourcing(SC) came into being. As an important part of SC, task assignment has received more and more attention. However, in the real scenario, the emergence of tasks is random and dynamic, which pose a huge challenge to task assignment. In order to solve this challenge, we propose a Dynamic Task Assignment Framework based on Prediction and Adaptive Batching (DTAF-PAB), which utilizes the Gated Recurrent Unit (GRU) in deep learning to predict the number of tasks entering a specific area, and propose an adaptive batching algorithm based on Deep Q Network (DQN) to dynamically adjust the size of batches, thereby improving the overall benefit of assignment. We use datasets from the real world to evaluate the competitiveness of DTAF-PAB and the experimental results show that the proposed framework is superior to other existing technologies in terms of both predictive performance and crowdsourcing platform benefit.

Coriolis flowmeters are designed based on the Coriolis force principle. It has high accuracy in mass flow measurement and provides additional multi-parameter measurements such as density, temperature, and volume flow. It is widely used in industrial flow measurement. Under the condition that the flowrate of single-phase flow is stable, the control and real-time tracking of most Coriolis flowmeters can meet the measurement requirements and obtain accurate real-time flow measurement results. However, when the flow conditions change rapidly, the converter system is often unable to track the flow changes in real time, and the measurement results may lag behind or even be erroneous. In this paper, a new prototype based on Zynq for the Coriolis flowmeter converter is developed in order to solve the problem of lagging behind in the measurement results. The fast sampling rate and more computational power from Zynq can realize the real-time measurement for rapidly changing flow. An experimental study using the prototype system was carried out under water flow conditions. Preliminary results showed that the repeatability under various test conditions down to one-second batch runs is within 0.1%. This verifies that the prototype system can measure rapidly changing flows.

Coriolis flowmeters rely on vibration to conduct their measurements, where the vibration is normally excited and detected by electromagnetic means. The excitation or detection system typically includes a coil and a permanent magnet to apply the excitation force or sense the vibration velocity on the measuring tube. This paper presents a multiphysics simulation method including both structural and electromagnetic finite element (FE) models together with a simplified single degree of freedom (SDOF) model to simulate the electromagnetic excitation and detection system. The finite element models of coil and permanent magnet were established through ANSYS Maxwell, which can represent either the excitation or the detection system. The measuring tube model of a U-shaped Coriolis mass flowmeter was established by ANSYS Mechanical, where the resonant frequency of the measuring tube and the frequency response at the detecting point were obtained. Combined with the established SDOF system model, the equivalent mass and equivalent stiffness of the measuring tube at the detecting point were also solved. By transferring the equivalent data of the SDOF system to the detection device established in Maxwell, the induced voltage by the detecting coil can be obtained. The experimental data of a single U-shaped Coriolis mass flowmeter (DN2) were collected for comparison with the simulation results. It has been demonstrated that the multiphysics simulation method can simulate the excitation force and induced voltage. The relative errors between the finite element calculation results and the experimental data are within 6%. This numerical method can be advantageously used to optimize the excitation and detection system design for Coriolis flowmeters.

Besides outliers generated from the data collection stage, anomalies in traffic data can also be valid data characterizing unusual traffic activities. Detecting these anomalies in spatiotemporal traffic activities can provide practical insights for traffic monitoring and operation. In this paper, we focus on detecting anomalies in spatiotemporal traffic demand data. We first apply a probabilistic tensor factorization framework to approximate the expected/predicted probability of each trip, and then quantify the degree of anomalous by using the log ratio of observed frequency over the expected probability. In this framework, each trip is considered a sample from an underlying multivariate categorical distribution of time of day, origin zone, destination zone, and day of week. We approximate this distribution using a probabilistic Tucker decomposition model and introduce an efficient expectation maximization (EM) algorithm for model inference. To test this framework, we design and implement two synthetic experiments in the traffic simulation software SUMO. The results show that the proposed framework can effectively detect anomalous activities in multivariate spatiotemporal traffic demand data.

Subway systems of most metropolitan cities worldwide are suffering from the problem of overloaded passenger flow especially during rush-hour of weekday. To address it, passenger flow control strategy is widely adopted by developing subway system in order to improve the operational efficiency and ensure passengers' safety. However, current strategy is usually formulated by the subjective experience of operation staff at each station. Their inability to grasp the correlation between station flow and link flow in the entire subway network can possibly cause reverse effect and passengers' dissatisfaction. In this paper, a new coordinated-based passenger flow control method, using the internal relation of stations and links, is proposed to handle this issue during rush-hour of regular weekday with the aim of minimizing the negative impact on irrelevant travelers. Using this approach, one can determine the target control stations with specific control strength in different period of rush-hour. An experiment is conducted on a realworld subway network in Guangzhou to examine the validity of the method. Result shows that our strategy is more advanced and effective compared to the actual one.

Gas flow facilities are essential to test gas flow sensors, meters and other devices. Pressure control is a key problem in particular for high pressure gas flow facilities. An appropriate control scheme is critical to improve the stability and the speed of the pressure control. This paper presents a linear Active Disturbance Rejection Control (ADRC) scheme for high pressure gas flow facilities with closed-loop pipelines. The control model of the flow facility is identified by a step response method. A reduced-order linear expansion state observer and a related control law are designed to estimate and compensate the total disturbance of the system. The robustness of the ADRC algorithm is compared with a PID control algorithm by Monte Carlo simulation. Control speed and disturbance rejection ability of ADRC are also compared with PID through simulation and gas flow experiments. The results show that, compared with PID, ADRC has a shorter settling time, stronger disturbance rejecting capability and more robust performance. The improved pressure stability has been proved to be beneficial to test gas flow sensors in the facility.

Recent years conventional neural network(CNN) has been applied to different natural language processing(NLP) tasks such as sentence classification, sentence modeling, etc. Some researchers use CNN to do multi-label classification but their work mainly focus on image rather than text. In this paper, we propose an improved CNN via hierarchical dirichlet process(HDP) model to deal with the multi-label classification problem in NLP. We first apply an HDP model to discard some words which are less important semantically. Then we use word embedding methods to transform words to vectors. Finally, we train CNN based on word vectors. Experimental results demonstrate that our method is superior to most traditional multi-label classification methods and TextCNN in terms of performance.

Under two-phase flow conditions, the performance of Coriolis mass flowmeters is significantly affected. A critical signal processing technique in Coriolis flowmeters is the estimation of resonance frequency and phase difference from sinusoidal vibration signals. This paper reviews three typical digital signal processing methods used to calculate phase difference, including Quadrature Demodulation, Hilbert Transform and Discrete Fourier Transform (DFT). These three methods are compared using simulated signals based on experimental results under two-phase flow conditions up to high GVFs (gas volume fractions) and low flow rate. The results indicate that, under two-phase flow conditions in particular for high GVFs and low flow rate, the three methods are affected in different degrees, and may need further modification to achieve better results under two-phase flow conditions.

Cache memory helps in expediting the speed of data retrieval time in processors in heterogeneous multi-core architecture, which is the main factor that affects system performance and power consumption. The implementation algorithm of cache replacement in current heterogeneous multi-core environment is thread-blinded, leading to a lower utilization of the cache. In fact, each of the CPU and GPU applications has its own characteristics, where CPU is responsible for the implementation of tasks and serial logic control, while GPU has a great advantage in parallel computing, which causes the need of cache blocks for CPU more sensitive than those for GPU. With that in mind, this research gives full consideration to the increment of thread priority in the cache replacement algorithm and takes a novel strategy to improve the work efficiency of last-level-cache (LLC), where the CPU and GPU applications share LLC dynamically and not in an absolutely fair status. Furthermore, our methodology switches policies between the LRU (Least Recently Used) and LFU (Least Frequently Used) effectively by comparing the number of cache misses on the LLC, which takes both the time and frequency of the accessing cache block into consideration. The experimental results indicate that this optimization method can effectively improve system performance.

Understanding passenger flow assignment patterns in a complex metro network is crucial to maintaining service reliability and developing efficient response during disruption. In reality passengers' perception of different cost attributes may vary with time. This paper focuses on quantifying the temporal variation of passenger route choice behavior and its impact on overall passenger flow assignment. In order to efficiently estimate model parameters, we modify a previous model to a missing data problem by introducing latent variable on route choice outcomes for each travel time observation. The revised model can be estimated using the Expectation-Maximization (EM) algorithm. We apply the proposed framework on Singapore's metro system and temporal grouped smart card transactions. We find that route choice coefficients vary substantially with time. The relative value of transfer time in terms of in-vehicle time ranges from 2 to 3, being higher at off-peak hours than during morning/evening peaks. The result suggests that passenger care more about total travel time during peak hours, whereas comfort (e.g., less transfer time) is of more concern to users during off-peaks. The proposed framework is general and can be applied on other networks.

This paper presents a design procedure for two-dimensional (2-D) zero-phase IIR notch filters based on the state-space representation of 2-D frequency transformation in the Fornasini-Marchesini second (FM II) model. It will be shown, via several typical image processing examples, that 2-D zero-phase IIR notch filters can be designed more flexibly by utilizing the state-space representation of the 2-D frequency transformation proposed by the authors recently, so that the distortion caused by nonlinear phase of the filter can be avoided and thus the designed filters perform better than the corresponding frequency domain filtering methods.

This paper extends the symbolic implementation of the new elementary operation approach (NEOA) to the multi-input and multi-output (MIMO) multidimensional (n-D) case, which is applicable to the causal transfer matrix with symbolic as well as numerical coefficients. Compared with the existing LFR Toolbox, the new developed package can generate the Roesser state-space model with a lower order, and the construction is direct without any pre- or post-processing. Moreover, the new version package supports the new symbolic engine MuPAD, which has been integrated in MATLAB since version 7.7. Symbolic examples will be given to illustrate effectiveness and facilities of the developed package.

—The objective of this research is to optimize the cross-sectional shape of an Averaging Pitot Tube (APT) flow sensor with Flow Conditioning Wing (FCW). The flat cutting and circle cutting wingtip shapes were researched through Computational Fluid Dynamics (CFD) simulation and air flow experimental tests for two different cross-sectional shapes of the APT flow sensor. The results show that the flat cutting wingtip shapes are better than the circle cutting, the wingtip width or wingtip arc radius of 0.3 mm are better than 0.5 mm, and the circumscribed circle diameter 8 mm is better than 6 mm as the cross-sectional shape is the same. Flow field analysis indicates that the wingtip shape determines the position of fluid flow separation, so affects the shape of the vortex area downstream of the detecting tube. The ratio of vortex width and length can express the vortex shape. The experimental data show that the smaller the ratio, the better the linearity and the repeatability of the sensor.

In this paper, an Averaging Pitot Tube (APT) with Flow Conditioning Wing (FCW) geometry is used as a practical sensing device to measure and characterize the flow of single-phase gaseous CO 2 . The technique demonstrates a simple, cost-effective and potentially accurate option towards the measurement, accurate accounting and characterization of CO 2 in Carbon Capture and Storage pipelines. The metrological performance of the flow sensor is verified using air medium before being applied for gaseous CO 2 . With a Coriolis mass flow meter acting as a secondary calibration reference to further validate the performance of the APT-FCW flow sensor, both metering instruments were evaluated against a weighing scale apparatus. From experimental and calibration data with air, the APT-FCW's average K-factor and linearity error are found to be 0.5091 and 0.725%, respectively. With operating conditions remaining unaltered in this particular flow measurement application and a target accuracy of ±1%, the error achieved for the Coriolis meter is better than ±0.5% and ±1% for the APT-FCW. Test results and other performance evaluation of the instruments are also discussed.

In many RFID applications, we conduct a RFID reader network to monitor a large number of tagged items. We also want to utilize this network to trace un-tagged items. In this paper, we propose ARNS-Adaptive RFID Network Scheduling scheme for device-free tracking. In particular, we firstly deploy a set of readers and reference tags into the monitored area. Then, we divide the tags within the read region of a reader into two groups, one in reader-to-reader collision area while the other not. We monitor the area by normal sampling model in which we need two phases to separately query the two groups. When an un-tagged target appears in the monitored area, we adopt tracking sampling model. In tracking sampling mode, the reader who detects the tags' frequency variations will adaptively query all tags in its read range to obtain the target's full trace. Meanwhile, we fill the missing path using shortest-path algorithm to recovery the trajectory. Experimental results show that ARNS achieves high tracking accuracy for certain applications.

In this paper, we analyze the performance of TID2008 and LIVE databases for full-reference image quality assessment metrics. A comparative analysis is presented. The subjective testing method which is used to obtain the perceived quality scores affects the accuracy and reliability of the database. It is shown that there are the relationship between the reference images and the types of distortions with different distorted levels. Results are: 1) Both TID2008 and LIVE databases are not well for image subjective quality assessment. 2)The distributions of their parameters are not adequate andreasonable. 3)The subjective methodology should use both the reference image and the worst distortion image.

This paper presents a new approach for underwater 3D target positioning by inhomogeneous illumination based on binocular stereo vision. Based on the theory of inhomogeneous illumination field, we build underwater binocular stereo vision system to measure the distance between the target and the detector (CCD receiver) as well as the 3D information of the target, which can be used to the further 3D reconstruction of underwater scene. The underwater experiments show that our system can get the depth as well as posture information of the targets accurately, especially in turbid water.

This paper mainly studies the design and optimization of diverter's nozzle in the flow facility. The profile of the flow at the outlet of the nozzle has great influence to the performance of the diverter. In case to optimize the diverter, we need to optimize the structure of the nozzle. As to the nozzle, an approximately symmetrical velocity profile across the jet width is desired. Here, we use the Fluent software combine with the experiment as tools to research the best structure of the nozzle.

A water flow diverter is an essential and error determining component of a liquid flow facility based on a static gravimetric system with flying start and finish. A new diverter was developed in order to reduce the uncertainty, noise and vibration by using a stepping motor to drive the dividing body. The reasons of the diverter timing error and test methods were studied. It is shown by the test results that the uncertainty of the diverter is better than 0.007%.

The paper deals with working principle and design methods about gas flow standard facilities by master meter method. A wind tunnel was designed to work as the flow velocity measurement and calibration platform. The distribution of the facility uncertainty was calculated. And finally the test results of the facility performance were given including the effect of the wind tunnel on the airflow, the impact of the humidity on the air density in the pipe and the flow-meter coefficient and the calibration of orifice flow-meter.

The driving behaviour of travellers has been found very different in case of emergency conditions compared to that at the normal traffic conditions. In this paper, we show how these different driving behaviors at the emergency condition, as well as the heterogeneity among drivers' behaviors, influence the network performance. We performed a sensitivity analysis on the model parameters representing the different (heterogeneous) driving behaviors and investigated the impact of these variations on the network performance. The analysis is conducted by applying a simulation framework using S-Paramics. The results show that reductions in the mean time headway and in the minimum gap dramatically increase the travel production (throughputs). Also, it is found that variations in driving behaviour play a much smaller role in the network performance.

In order to solve the problem that mechanical pointer type flowmeter cannot be calibrated automatically, based on the photoelectric conversion principle, a non-contact meter pointer automatic identification sensor was designed. The signal acquisition hardware and software of the pointer identification sensor were designed also. Two flowmeters can be calibrated at the same time. It was shown by the result of experiments that automatic recognition of meter pointer, automatic signal acquisition and processing were achieved. It meets the industrial field demand of mechanical pointer type flowmeter automatic calibration.

The differential pressure exported by the averaging Pitot tube is small, it requires an expensive micro-scale differential pressure transducer with very good performance. In order to overcome this drawback, a rotameter was used to take place of the differential pressure transducer in the paper. The rotameter and the main pipe are in parallel lines, so that the rotameter is a by-pass pipe. The scale of the rotameter can reflect the flow rate of the main pipe indirectly. A mathematical model can be gotten according to the experiment, and then linear fitted the curve of the flow rate, so that an easier mathematical model can be obtained. In three different conditions, the method of the averaging Pitot tube with rotameter and the linear fitted model were verified. It is shown in the experiment that, the relative error between the calculated flow rate and the actual flow rate can reach 2.79%. If the requirement of the accuracy is low, this method can satisfy the requirements in the industrial fields.

In this paper, the evaluation methods of flow standard facility in every regulations were compared and analyzed. Analysis of the two performance parameters of the flow stability within cumulative time and the flow stability between cumulative time. From the two angles of the structure of turbine flow meter itself and the calculation formula of the stability parameter, the selection principle of the stability parameter is determined. The flow stability between cumulative time is selected as the main parameter, and the flow stability within cumulative time as reference. And the results of flow stability test between the method of turbine flow meter and the weighing method were compared. The results show that the method of turbine flow meter can also be an accurate method for testing the flow stability of flow standard facility.

With the turbine flow sensor applications increasingly widespread, the requirements of the turbine flow sensor accuracy are also higher. In this paper, the simulation model was built in strict accordance with the experimental 15mm diameter turbine flow sensor's geometric structure with careful grid resolutions. Five kinds of turbulence models, including standard k-ω, SST k-ω, RNG k-e, Realizable k-e and RSM, were used in three-dimensional simulation of 15mm diameter liquid turbine flow sensor and compared with the experimental data. The results showed that the relative error of the average meter factor between simulation and experimental results can reach 2.6% level. RNG k-e and RSM model can more exactly predict the sensor's linearity. Standard k-ω can be more exact for the prediction of the meter factor of the sensor. The friction torque between the shaft and bearings obtained by theoretical formula has little effect on the turbine flow sensor simulation accuracy.