Search

Images taken in extremely low light suffer from various problems such as heavy noise, blur, and color distortion. Assuming the low-light images contain a good representation of the scene content, current enhancement methods focus on finding a suitable illumination adjustment but often fail to deal with heavy noise and color distortion. Recently, some works try to suppress noise and reconstruct low-light images from raw data. But these works apply a network instead of an image signal processing pipeline (ISP) to map the raw data to enhanced results which leads to heavy learning burden for the network and get unsatisfactory results. In order to remove heavy noise, correct color bias and enhance details more effectively, we propose a two-stage Low Light Image Signal Processing Network named LLISP. The design of our network is inspired by the traditional ISP: processing the images in multiple stages according to the attributes of different tasks. In the first stage, a simple denoising module is introduced to reduce heavy noise. In the second stage, we propose a two-branch network to reconstruct the low-light images and enhance texture details. One branch aims at correcting color distortion and restoring image content, while another branch focuses on recovering realistic texture. Experimental results demonstrate that the proposed method can reconstruct high-quality images from low-light raw data and replace the traditional ISP.

Since the introduction of mica paper and synthetic resin-based taping insulation for rotating-machines in 1950s, only evolutional improvements in their properties and processing have been made. Due to its challenging nature, much of recent insulation research has been focused on the aspect of reliability rather than performance enhancement. This paper presents the development and performance evaluation of a revolutionary nanostructured insulation in the manufacturing of large propulsion motors with game-changing torque density and payload efficiency for marine Next Generation Integrated Power System. It is demonstrated that nanostructured insulation based on 2D-platelet fillers could offer significant improvement over conventional insulation system in electrical, dielectric, thermal and mechanical properties. Optimal nanostructured insulation formation identified through a Design of Experiment study exhibits high thermal conductivity of $>0.8$ W/( $\text{m}\cdot \text{K}$ ), high breakdown strength of >40 kV/mm, low dielectric constant of less than 5.5 and low dielectric loss factor of less than 2.5% at 155°C. Voltage endurance tests on coupons with optimal formulation, in accordance with IEC 60343 standard, demonstrate satisfactory endurance life. Furthermore, a multi-physics finite-element-analysis model of the winding in a medium-voltage electric motor is established to perform electromagnetic and thermal analyses in ANSYS under various temperature boundary conditions. The study indicates a remarkable increase of 14% in torque handling capacity of a motor wound with proposed nanostructured insulation material when compared to an identical motor wound with conventional micaceous insulation.

This paper studies the distributed state estimation problem for a multi-sensor network where each sensing node has different process model which means that the state estimation cannot be fused directly with other nodes’ estimation states. When the discrete time-varying process model and affine transformation between two estimation spaces of multi-sensor network are a prior information, to get a consistent estimation set of target, the transformed process model of each sensing nodes is derived, and a consensus based distributed Kalman filtering algorithm via indirect state vector fusion (SVF) strategy is designed. If the communication topology of network is connected, the estimation errors of sensors are globally asymptotically stable without the process and measurement noises. In the existence of the process and measurement noises, it is proven that the estimation errors are exponentially bounded in mean square and bounded with probability 1. And the boundary of inverse covariance error matrix and Kalman gain are derived in the meantime.

Applications of local input-to-output stability theorem mainly focus on isolated power systems. This paper aims at giving a try to extend the local input-to-output stability based stability criterion to integrated energy system, narrowing down on how to construct the linear gain matrix in the interconnected energy system. By modify multiregional inoperability input-output model, a formula of the linear gain matrix is derived, with a simple energy system as the case study.

The Free Viewpoint Video (FVV) system is an advanced visual media mode, which could satisfy the audience's demand of watching from any viewpoints. It has aroused great interests in the field of multimedia and vision. This paper proposes a set of solutions, including depth estimation, virtual view interpolation and post processing, to synthesize and optimize the virtual view images. Specifically, deep learning based method is proposed to compute depth images, which is more efficient than traditional stereo matching algorithm. The virtual view can be synthesized with the help of DIBR algorithm in an arbitrary position between two reference cameras. To remove the noise caused by rendering, some post processing techniques, such as median filtering and hole filling, are employed. The extensive experimental results demonstrate that the quality of the synthesized free viewpoint video has been significantly improved using our method.

With the emergence of massive multimedia data, how to securely and effectively manage them in smart homes has become a crucial task. Although the digital photo frame can satisfy the display function, the current digital photo frame system cannot provide users with personalized services. Therefore, in this paper, we propose a digital photo frame system with online hashing image retrieval. The system is mainly divided into two parts: First, we leverage the image retrieval based on online hashing to manage photos. When a user selects a photo, we retrieve it and return the similar photos with background music based on the photo semantic information. Then, in order to enhance system security, we build the privacy mode, so that users can manage the data in security. At the end, we build a dataset for smart homes, and implement a multifunctional secure digital photo frame system.

Speaker recognition technology has been widely used in home automation recently, which can provide identification for the user efficiently. However, speaker recognition technology brings security risks as well as convenience, especially the privacy disclosure of the user's voiceprint templates. This paper proposes an efficient and privacy-preserving speaker verification scheme for home automation devices, named SEASON. We modify the high-accuracy deepspeaker framework in terms of privacy, based on residual network. With SEASON, users can access the home automation devices securely. Detailed security analysis shows that SEASON can actually resist various security threats. We implement SEASON with real datasets and 48 real records from four testers. Experiments reveal that SEASON can achieve efficient and accurate verification.

The baseline profile of the third generation of Audio Video coding Standard (AVS3) was finalized in March 2019, and the high profile is expected to be issued in 2021. In the AVS3 baseline profile, various key coding technologies on coding structure, prediction, transform, and loop filter are newly adopted. AVS3 baseline profile achieves a significant improvement in coding efficiency relative to the previous video coding standards. The experimental results show that the bitrate savings are 23.52% and 22.25% averagely compared with AVS2 and HEVC, respectively. The computational complexity of AVS3 rises inevitably due to the newly introduced coding tools. This paper highlights the new technical features in AVS3 and analyzes the coding performance and computational complexity of the principal coding tools in AVS3 by providing specific experimental results.

In this paper, we present a novel approach for image translation based on Generative Adversarial Networks (GANs). We apply the self-attention mechanism to improve the quality of the generated images as well as focus on not only local feature representation but also global structural correlation. Meanwhile, we adopt the idea of cyclic image translation in CycleGAN. Moreover, in order to stabilize the training process and reduce the probability of appearing unusual gradients, two techniques spectral normalization and TTUR applied during training process. In the experimental section, we compare our method with the original CycleGAN in terms of subjective evaluation, objective scoring, classification verification and computational cost. Both subjective and objective results show that our model can generate high quality and abundant diversity images that uses unpaired unlabeled samples. Two techniques used during the training have prompt the convergence speed of the network and help the model use less time to get the optimal point. Lots of comparisons against CycleGAN demonstrate that our proposed method is superior to the original one.

Multi-level inverters (MLIs) have shown popularity among medium voltage marine drives, especially as they allow for increased reliability through reconfiguration. Many of today’s voltage source inverters (VSI) utilize space vector pulse width modulation (SVPWM) compared to sinusoidal PWM (SPWM) due to the advantages of SVPWM in providing increased modulation index. Today’s MLIs are limited in the number of levels for various reasons. This includes the ability to stack a large number of switches and coordinate their turn on/off transients, availability of control platforms that provide enough PWM channels, and ability to produce large space vector maps. This paper’s main focus is to create a method for direct calculation of the switching sequence for neutral point clamped (NPC) inverters, for any Nlevel inverter, which is essential when MLIs are reconfigured for different levels to mitigate semiconductor switch faults. Calculation of the desired vertices to achieve a certain voltage vector is purposed through a novel projection and mapping of the reference vector in the space vector diagram.

This paper investigates the effect of impact factors (IFs) to harmonic content of an electric machine drive circuit, post-fault reconfigured multilevel inverters (MLIs) by using an accurate mathematical model, and impact of reconfiguration methods on THD. Due to the difficulty of analyzing harmonic content of MLIs mathematically, software tools are typically used to simply calculate the value of THD as metric. In this paper, an accurate mathematical model including both low- and high-frequency contents of harmonics in MLIs is utilized to derive the general expression of their output voltage THD. The effect of IFs on harmonic content are analyzed based on the general expression. Simulations and experiments are performed to validate the proposed mathematical model. The impacts of the reconfiguration methods on THD are analyzed. The impact of THD and other essential metrics on power quality are analyzed, and an minimal degradation of power quality caused by reconfiguration methods is proposed.

In electric drives, semiconductor devices are susceptible to faults due to high switching stresses. An unsupervised principal component analysis (PCA) approach that utilizes the on-state median values of voltage and current signals as features is proposed; it is computationally efficient for realtime operation. A key innovation is handling variations of the operating point in the feature space using exponentially weighted samples to recursively compute the time evolution of means of on-state features used in a PCA-based anomaly detector. Data from accelerated tests on commercial SiC devices are used to validate the effectiveness of the proposed approach in predicting incipient semiconductor faults. An experimental setup is also used to demonstrate real-time implementation of the proposed method.

Rapid growing demand in higher power/torque density and payload efficiency of marine electrical propulsion can be addressed by a high heat transfer, indirect cooled stator insulation system. A novel nanocomposite insulation has been developed for high torque density machine through proper dispersion of 2D nanoclay in epoxy resin matrix to obtain improvement in both thermal and electrical performance. The voltage endurance testing shows that while neat epoxy resin failed within 500 hours, the nanocomposite survived for more than 3000 hours. The erosion rate, evaluated by the depth and volume, of nanocomposite was also significantly slower than the neat epoxy resin. The layered structure of clay is attributed to benefit the discharge resistance.

Most of the present high-voltage high-frequency insulation material testing focuses on the thermal and electrical properties of transformer oil and cable insulation. This paper elucidates the methodology and the experimental setup for high-voltage high-frequency testing of motor winding insulation. The experimental setup gives a pulsed output of voltages ranging from 2 kV till 10 kV peak-to-peak. The experimental setup is used to test the insulation material Kapton HN at different thicknesses [1mil, 2mil, 3mil and 5 mil) at a frequency of 3kHz. Results of the breakdown tests as well as the experimental setup are discussed in this paper.

The main focus of this paper is to propose a reconfiguration method to mitigate short circuit faults in neutral point clamped (NPC) multilevel inverters. Existing reconfiguration methods use either a redundant strategy or need additional devices pre-installed in the circuit which will increase the bulk and complexity of the system. Different from these methods, control-strategy-based reconfiguration is the most robust one since it does not require any change in the circuit configuration. However, most existing control-strategy-based methods distort the phase-to-neutral voltage which results in a degradation of power quality. To maintain control-strategy-based reconfiguration and avoid phase voltage distortion, a new and practical reconfiguration method is proposed in this paper. The proposed method is applicable to any NPC multilevel inverter. MATLAB/Simulink is used to simulate a five-level NPC inverter with non-idealities to verify the proposed reconfiguration method. A single-phase five-level NPC has been built in hardware, tested in healthy condition, and the proposed method has been validated in a low voltage condition.

The main focus of this paper is to investigate the effect of post-fault reconfiguration on the inverter voltage total harmonic distortion in multilevel inverters (MLI). Three different operating conditions are compared: 1) Healthy, 2) Faulty, and 3) Reconfigured using different methods that utilize inherent inverter redundancy. Literature mostly addresses balancing line-to-line voltage RMS when reconfiguring the inverter after fault occurrence, and line-to-neutral voltage and harmonic content after reconfiguration is ignored. However, line-to-neutral voltage quality is essential for per-phase analysis of Y-connected three-phase loads, neutral point power conditioning when the neutral exists, unbalanced loading conditions, and harmonic content analysis required by some grid interconnection standards such as AS4777 and IEEE1547. This paper first introduces an accurate analytical model to analyze the harmonic content of the line-to-neutral voltage. Then, two fault conditions are introduced into the model and verified in experiments. Three popular reconfiguration methods to mitigate these faults are then evaluated to analyze and compare harmonic content and RMS values of the achieved line-to-neutral voltages. Simulation and experimental results of healthy, faulty, and reconfigured conditions show that the predicted RMS and total harmonic distortion of the line-to-neutral voltage using the proposed model are accurate. Results also show that the line-to-neutral voltage is severely impacted by existing reconfiguration methods.

In this paper, a novel topology and control scheme for active electromagnetic interference (EMI) filter based on the idea of virtual impedance enhancement is proposed. To ease analysis, a general and simplified equivalent circuit is introduced to predict the conducted emissions incurred from switched-mode power converters. In addition, theoretical analysis is given to show how the proposed active EMI filter (AEF) should be designed along with simulation validation. A prototype of proposed AEF is implemented and tested to verify its effectiveness on proof-of-concept test setup. It is shown that the proposed AEF is capable of attenuating conducted emissions.

This article investigates the efficiency of a propulsion motor drive system due to load sharing. Two 2250HP-rated induction motor drives controlled via indirect field oriented control are simulated. Load sharing with different configurations is studied to analyze the impact of individual drive loads on the multi-drive system efficiency. Motor losses, Inverter losses and propeller losses are accounted for in efficiency calculations. For each load torque shared, the individual drive efficiency is calculated to determine which load sharing scheme provides maximum system efficiency. The approach for load sharing and system efficiency maximization is applicable to ship propulsion motor drives for maximum efficiency beyond rated load torques, and to other multi-drive systems.

This paper proposes a generalized logic-based method for fault diagnosis in multilevel inverters that can be integrated with reconfiguration methods. Thus, both effective fault diagnosis and fault tolerance can be accomplished. To validate such integration, the proposed fault diagnosis method is applied to a five-level neutral point clamped inverter with three reconfiguration methods proposed in the technical literature. The effectiveness of the fault diagnosis method is verified with simulation and experimental results, while the integration of this method with the reconfiguration techniques is validated with MATLAB/Simulink simulations. Several simulation and experimental results under healthy and faulty conditions are presented to demonstrate the satisfactory performance of the proposed integrated fault diagnosis and recovery.

It has been identified in ONR Next Generation Integrated Power System (NGIPS) Roadmap that fundamental research in dielectric insulation research enables payload efficiency and affordable high power density of integrated electric propulsion motors. The objective of this study is to revolutionize electrical insulation in the manufacturing of NGIPS motors for military electric propulsion with game-changing torque density and payload efficiency through the development of nanostructured insulation with significant improvements in both electrical and thermal performance. This paper presents the progress of nanostructured insulation innovation, high field characterization, performance and insulation integrity validation under high voltage, high frequency multi-stresses. The thermal, dielectric and voltage endurance properties of novel nanocomposite insulation based on 2D-nanostructured platelet fillers were investigated. It was demonstrated that nanostructured insulation could offer significant improvement over conventional insulation system in electrical, dielectric, thermal and mechanical properties. A Design of Experiment was employed to study the effects of various 2D fillers and their interplay, and more importantly to identify the optimal nanostructured formulation with high thermal conductivity of >0.8 W/mK, low dielectric constant of less than 5, low dissipation factor of less than 3% at 150°C and high breakdown strength of >1000V/mil. Furthermore, disk samples with optimal formulation from Design of Experiment were fabricated for voltage endurance tests in accordance with IEC 60343 standard for evaluation of their long-term endurance life.

This paper compares two main topologies used for capacitive-coupled consumer wireless charging which has several advantages over inductive coupling including lower cost and no shielding requirements. There has been minimal data in the literature on preferred transmitter topologies. Two types of transmission systems are compared in simulations and experiments. An experimental charging system was built and tested with the evaluation of efficiency of h-bridge and half bridge transmitters. Preliminary experimental results indicate that the h-bridge is both economically and efficiency-wise advantageous over the half bridge in this application. The overall results show that the H-bridge is more efficient even though it has more semiconductor switching devices causing it to have a slightly higher cost. Immediate future work is implementation of closed-loop control for receiver-side voltage regulation.

This paper presents a real-time integrated model of a micro-grid to simulate its electrical energy infrastructure. This infrastructure includes two PV arrays, a fuel cell, and a diesel generator that support building loads when islanded from the utility grid. The paper reviews existing models, which are usually available either as 1) low-level dynamic models, along with power electronics for specific components, e.g. PV system or fuel cell, or 2) high-level such as with conventional power systems where power electronics are ignored due to their faster dynamics. The proposed modeling strategy for sustainable power generation is emphasized for both grid-connected and islanding modes and combines slow and fast dynamics where both a micro-grid and related power electronics dynamics are simulated to show how a high-fidelity model can be used in a dynamic micro-grid environment. A synchronized regulator for islanded mode is presented. The PV arrays and fuel cell are assumed to be always available and variable irradiance conditions (e.g. nighttime) and the change of hydrogen and oxygen densities of fuel cell are shown in the paper. The diesel generator is used for black start or when the utility grid voltage or frequency drop below a threshold during which potential grid collapse could occur and the micro-grid goes into the island mode. The micro-grid model is simulated using a real-time simulator so that longer case studies and scenarios can be studied without ignoring fast dynamics. The main contribution of this paper is that both these time scales (slow and fast) are integrated in this real-time simulation platform for more realistic performance analysis. The results show the ability of the integrated micro-grid simulation to function in both grid-tied and islanded modes.

This paper presents a threshold-based approach for power grid fault diagnosis in a power system based on correlation between fault types and basic measurements. The main focus is on faults that may occur on transmission and distribution lines due to uncontrolled events. The proposed method utilizes simple voltage and current measurements in transmission lines to extract divergence of these measurements from typical or nominal values. The first phase of the method is similar to a “training” phase where thresholds are identified for each measurement, followed by the second phase where binary combinations of comparisons with thresholds are used to diagnose specific faults. The IEEE standard 14-bus system is used to illustrate how the proposed method can be applied for transmission line fault diagnosis. Two- and three-phase to ground faults in three transmission lines are injected in a simulation platform as an example, and their effect on different current measurements are observed. Two basic quantities are observed for each of the measured signals: 1) RMS value, and 2) harmonic content. Measured quantities are shown to accurately capture unique binary signatures of different faults so that they can be utilized in offline and/or online fault diagnosis.

This paper presents a fuzzy-logic approach for fault diagnosis in a plug-in hybrid electric vehicle (PHEV) or electric vehicle (EV) charger. The approach correlates faults and basic measurements then utilizes fuzzy logic to diagnose a fault. Redundancy is engaged as needed based on the fuzzy logic decision. Experimental validation is first shown for the converter model. Then, a fuzzy system is developed based on correlation of faults and measurements. Faults considered are open-circuit and short-circuit faults in major components. Results show that all faults are accurately detected even though two pairs of faults are confused and are thus both flagged. Design of the fault diagnosis method is shown to be simple and can be applied to a wide array of energy systems.

This paper presents a time-efficient modeling and simulation strategy for aggregated microinverters in large-scale simulations, with a 20-panel system used as an example. As the microinverter configuration emerges to overcome reliability and loss mismatch issues in centralized PV inverters, a model that can be used for rapid prototyping is required. This model is also helpful in predicting behaviors of individual microinverters in a large-scale system, e.g., power generation under partial shading conditions and harmonic interaction. Dynamic microinverter models that include switching action may not be suitable for simulating multiple microinverters as they yield a significant computational burden and long simulation time compared to simplified but accurate average models. This paper introduces the average model of a basic microinverter composed of a boost stage and an H-bridge represented by a single matrix form aggregation of multiple microinverters in a larger system to achieve faster simulation times with accurate transient and steady-state results. A 200W prototype microinverter board is tested to verify the proposed average and dynamic models. Furthermore, MATLAB/Simulink is used to show the improved simulation speed and accuracy in the multiple-microinverter configuration.

This paper presents a generalized approach for intelligent fault detection in power electronic converters based on correlation between faults and basic measurements. Fault recovery is then applied based on this correlation by using specific signals and quantities from existing measurements. The main purpose is for power electronic systems to cope with the notion of a smart grid and its different smart components. Existing intelligent control of power electronic systems are first overviewed. Then, short- and open-circuit faults in each power electronic component are injected in a simulation platform, and their effect on different voltage and current measurements are observed. A simple block with controlled short- and open-circuit operation is developed for this purpose. Three basic quantities are observed for each of the measured signals: 1) average value, 2) RMS value, and 3) harmonic content. Both simple logic and intelligent control are used to engage redundant components as a first step to fault recovery. A solar photovoltaic (PV) micro-inverter in stand-alone mode is used as a testing platform for this paper. Simple logic based control is presented here and shows the ability to detect several faults in the power stage. Fuzzy logic is also introduced in such an application using an example.

Due to compression sensing wide research, sparse representation was well applied in super-resolution. At present, super-resolution via sparse representation is performed on the all patches of the image, but it is high time-consuming and its results are not perfect sometime. Considering the difference in the structure of the image patches, we proposed a super-resolution based on image patches structure feature. The method classifies the patches into smooth region and non-smooth region based on gray variance of the patch, and high-resolution and low-resolution dictionaries, for super-resolution via sparse representation, are obtained based on patches structure, and then bicubic interpolation and super-resolution via sparse representation are used to smooth region and non-smooth region respectively. Experiments show that this method can obtain better results and the run-time of image super-resolution is reduced comparison with the Yang et al. [9] method.

In this paper, a novel resource management approach has been present to increase the resource utility of the heterogonous network under the IEEE1900.4 architecture. Based on the prediction of the handoff call arrival rate and the variation of traffic in different hours, system resource has been reserved dynamically and re-allocated for each RAN, so that the performance of systems is increased.

In the present work, microswitches with different shapes have been fabricated by low-cost and convenient multi-layer electroplating in order to develop a shock acceleration microswitch with enhanced-contact and low off-axis sensitivity. Modal analysis based on FEM shows that the three kinds of new designed microswitches have lower off-axis sensitivity than our previous device, i.e., type I. The packaged microswitches were tested by the drop hammer system. The generated half-sine-like shocking acceleration with amplitude of 80g, lager than the threshold, was applied to the microswitches. The test contact time of the microswitch IV is about 55µs, whose contact effect is much better than conventional. A shock acceleration microswitch IV with a movable contact point utilizes the double spring-mass system to realize an enhanced-contact effect and is considered as a better selection for long duration contact and relatively low off-axis sensitivity.

A broadband unidirectional coplanar waveguide (CPW) fed rectangular slot millimeter wave antenna based on micro-electro-mechanical systems (MEMS) non-silicon micro-fabrication technology is discussed. A wide impedance bandwidth of 75% (Su