Search

Polar Harmonic Transform (PHT) is termed to represent a set of transforms those kernels are basic waves and harmonic in nature, which can improve the effect in Intelligent Transportation System (ITS) applications. PHTs consist of Polar Complex Exponential Transform (PCET), Polar Cosine Transform (PCT) and Polar Sine Transform (PST). PHTs can extract orthogonal and rotation invariant features and demonstrated superior performance in various image processing and computer vision applications. For real time systems and large multimedia databases, execution efficiency is always a significant challenge. With widespread use of Graphics Processing Unit (GPU), this study presents GPU based PHTs. Proposed methods are based on mathematical properties of PHTs and optimization techniques of GPU. Optimal parameter selections for GPU execution are also discussed. In our experiments, proposed methods are over 1800 times faster.

In this paper, we propose a point cloud based 3D object detection framework that accounts for both contextual and local information by leveraging multi-receptive field pillars, named as MuRF-Net. Recently, common pipelines can be divided into a voxel-based feature encoder and an object detector. During the feature encoding steps, contextual information is neglected, which is critical for the 3D object detection task. Thus, the encoded features are not suitable to input to the subsequent object detector. To address this challenge, we propose the MuRF-Net with a multi-receptive field voxelization mechanism to capture both contextual and local information. After the voxelization, the voxelized points (pillars) are processed by a feature encoder, and a channel-wise feature reconfiguration module is proposed to combine the features with different receptive fields using a lateral enhanced fusion network. In addition, to handle the increase of memory and computational cost brought by multi-receptive field voxelization, a dynamic voxel encoder is applied taking advantage of the sparseness of the point cloud. Experiments on the KITTI benchmark for both 3D object and Bird's Eye View (BEV) detection tasks on car class are conducted and MuRF-Net achieved the state-of-the-art results compared with other voxel-based methods. Besides, the MuRF-Net can achieve nearly real-time speed with 20Hz.

Image inpainting, also known as image completion, is the process of filling in the missing region of an incomplete image to make the repaired image visually plausible. Strided convolutional layer learns high-level representations while reducing the computational complexity, but fails to preserve existing detail from the original images (eg, texture, sharp transients), therefore it degrades the generative model in image inpainting task. To reduce the erosion of high-resolution components of images meanwhile maintaining the semantic representation, this paper designs a brand-new network called Interactive Separation Network that progressively decomposites the features into two streams and fuses them. Besides, the rationality of network design and the efficiency of proposed network is demonstrated in the ablation study. To the best of our knowledge, the experimental results of proposed method are superior to state-of-the-art inpainting approaches.

Polar Fourier analysis can extract orthogonal and rotation invariant features and demonstrate superior performances in many image processing tasks. For real world applications, execution efficiency is always a significant challenge. With widespread use of Graphics Processing Unit (GPU), this research presents GPU accelerated polar Fourier analysis. Proposed parallel algorithm is based on mathematical properties of polar Fourier analysis and optimization techniques of GPU. Optimal parameter selections for GPU execution are also evaluated. In our experiments, with same computation result proposed method is over 170 times faster. Wide range of applications that using polar Fourier analysis will inspired from this work.

Irregular text refers to text with incomplete sequence information of the sentence or text that does not meet normal grammatical specifications, such as Weibo text. Existing named entity recognition algorithms recognize such text, the effect is poor due to the lack of context information. Because the attention mechanism has advantages in obtaining contextual information, we merge part-of-speech attention with the BI-LSTM-CRF model and propose a BLAC model. We tested on several public datasets and compared the results with the basic model BI-LSTM-CRF. The results show that the method we proposed has a certain improvement in the entity recognition of irregular short text.

Mobile edge computing (MEC) is a promising technology to provide computing services for resource-constrained mobile users, improving the computing experience. However, energy consumption is also a considerable expense for users. In this paper, we investigate the joint wireless resource allocation, power control and computation offloading for an MEC-based multi-user wireless communication system. The objective of this paper is to minimize the energy consumption of all users. We formulate the problem as an optimization problem and propose to solve the optimal power control strategy by using convex optimization method. Then we derive the semi-closed forms of resource allocation and computation offloading. By using dichotomy algorithm, we have obtained the resource allocation and computation offloading policy. Simulation results show that the proposed scheme effectively reduce the user energy consumption.

Eye tracking is a valuable topic in computer vision. Appearance-based eye tracking is a promising research direction in recent years. Convolutional neural networks (CNN) had been used in gaze estimation, which cover the significant variability in eye appearance caused by unconstrained head motion. With computation capability of consumer devices rapidly evolving, accurate and efficient appearance-based eye tracking has the potential for multipurpose applications. Person-independent networks have limit in improving gaze estimation accuracy. Person-specific network with calibration is more effective than person-independent approaches. Unlike classical eye tracking methods, appearance-based eye tracking has not a clear way to calibration. Our goal is to analyze the impact of calibration data selection and calibration target distribution on person-specific gaze estimation accuracy. We trained person-independent network and use SVR to calibration. We choose two kind of typical distribution targets to evaluation. Use different distribution targets to calibration achieves different accuracy.

Angular radial transform (ART) is a shape descriptor in MPEG-7 for image representation and image analysis. For real world applications, execution efficiency is always a significant challenge. With widespread use of Graphics Processing Unit (GPU) in image processing tasks, this paper presents GPU based angular radial transform (GART). Proposed parallel GART is based on mathematical properties of ART and optimization techniques of GPU. Optimal parameter selections for GPU execution are also evaluated. In our experiments, with same computation result GART is over 1800 times faster than ART. Wide range of applications that using ART will inspired from this work.

In this paper, a high-gain circularly polarized (CP) antenna array operating at Q-band is presented. The antenna element is composed of a radiating tilted cross slot etched on the top aluminum layer and an substrate integrated waveguide (SIW) coupling slot on the bottom substrate layer. CP radiation is excited by the tilted cross slot. A 1 × 4 array with a low loss SIW feeding network is designed based on the proposed antenna element. Simulated results show that the proposed antenna array features the advantages such as high gain and wide bandwidth.

A novel fault-tolerant inverter applied in this islanded microgrid is proposed in this paper. Firstly, the topology structure and the working principle of the microgrid with the fault-tolerant inverter are provided. The fault-tolerant inverter allows the switching devices of different bridge arms to be reconstructed, which improves the fault-tolerant space. A fault-tolerant pulse signal algorithm is proposed, combined with the droop three-loop control to form a control method for the fault-tolerant inverter. It is no need to switch the control algorithms between the normal operating mode and the fault-tolerant mode. The reliability of the system is improved. The circulating currents eliminating method between the two inverters when the fault-tolerant inverter is in fault state are analyzed. Finally, the validity of the topology and the control strategy of the system is verified by simulation.

In order to realize the application of ultrasonic wave in the internal structure detection of transformer, an ultrasonic transmitting and receiving circuit based on STM32 is designed. The transmitting circuit includes an optocupler isolation circuit, a MOSFET driving circuit, a push-pull conversion circuit and an impedance matching circuit, etc., and generates a high-voltage excitation pulse. The receiving circuit includes an amplitude limiting circuit, a preamplifying circuit, a band pass filtering circuit and an automatic gain control circuit to implement processing of the ultrasonic echo signal. The actual test shows that the ultrasonic excitation signal generated by the system is small in noise, steep in waveform, clear in echo signal after processing, good in system performance, simple in structure, stable and reliable, and has good promotion value.

An online calculation method for emerging decision-making of static voltage stability based on severe fault set is proposed, which is of great significance for ensuring the safe operation of the power grid. The method aims to select the reactive power control measure based on the voltage modal analysis result; when the reactive power control is insufficient, the active power control measure is added based on the active comprehensive sensitivity information; when the fault comprehensive control measure causes the normal operation mode When the bus voltage is upper limit, the reactive power distribution is re-adjusted by solving the reactive voltage comprehensive sensitivity. The application effect in Liaoning network has proved the practicality of assistant decision-making.

Liquid metal current limiter (LMCL) based on forced current commutation is potential to be used in MVDC power system. This LMCL utilizes fast switch (FS) and a parallel connecting path consists of several current limiting elements (CLEs) and an auxiliary magnetic induction module (MIM) in serial. The MIM consists of a pre-charged capacitor C, a thyristor T and a set of magnetic coupling coils (MCC). In this paper, a simulation model of this LMCL is built to investigate the current limiting process. And it is experimentally verified according to current commutation and limitation characteristics. Furthermore, based on this model, the current commutation characteristics are analyzed allowing for different influencing factors including the stray inductance of CLEs path and the parameters of MIM. It can be concluded that the effect of CLEs path stray inductance plays a vital role to current commutation process. Finally, the pre-arcing characteristics of CLE during current commutation process are experimental revealed. The pre-arcing duration tends to decrease with the first 100μs current rising rate (di/dt) increasing. And the arc will not ignite during current commutation when the first 100μs di/dt is 200A/μs at specific conditions.

A topology of fault-tolerant inverter with large fault-tolerant space is proposed in this paper. It can be applied to the battery storage. The inverter circuit needs to be reconstructed to keep system running normally when the power switches are faulty. The control strategy is proposed. The fault-tolerant inverter doesn't need to convert the control algorithm from the normal operation state to the fault operation state. The system reliability is improved. The system stability is analyzed. Simulation results verify the correctness of the topology of fault-tolerant inverter and the control strategy for battery storage in islanded microgrid.

The power system plays a major role in the maintenance of working properly for satellite. As there are many working loads and different working attitudes, the power system has many diverse working patterns. So it is very critical to recognize the working patterns accurately. However, due to the measurement error, environmental interference, and other uncertainty factors, the output voltage of the satellite power system has remarkable uncertainty. If we did not consider the uncertainty and various working patterns, poor recognized result will be generated. For this issue, we proposed a working patterns recognition method for satellite power system based on uncertainty data clustering strategy. In this method, we firstly utilize uncertainty data clustering strategy to modeling working patterns. Then during pattern recognition stage, we calculate the distances between uncertain cluster centers and the measurement data. The experimental results of actual power system data illustrate the validation and feasibility of our proposed method.

Ambient Occlusion (AO) is a technique to approximate the effect of environment lighting and add realism to a scene by accentuating surface details and adding soft shadows, which is widely used in multimedia applications. Neural Network Ambient Occlusion (NNAO) is a pioneer in introducing deep learning to accurate and real-time AO, but it has two limitations: 1) performance bottleneck under excessive amount of samples ; 2) low contrast and blurred edges leading to unreal effect. To overcome these two limitations, we propose Accelerated Detail-enhanced Ambient Occlusion (ADAO) method based on NNAO by adopting three image processing methods: 1) spiral sampling in screen space; 2) contrast enhancement of AO map; 3) normal-depth edge preserving bilateral filtering. Experimental results show that the proposed method is over 2 times faster than NNAO and produces shadows with more realistic details.

Consistent melt pool geometry is an indicator of a stable laser powder bed fusion (L-PBF) additive manufacturing process. Melt pool size and shape reflect the impact of process parameters and scanning path on the interaction between the laser and the powder material, the phase change and the flow dynamics of the material during the process. Current L-PBF processes are operated based on predetermined toolpaths and processing parameters and consequently lack the ability to make reactions to unexpected melt pool changes. This paper investigated how melt pool can be characterized in real-time for feedback control. A deep learning-based melt pool classification method is developed to analyze melt pool size both fast and accurately. The classifier, based on a convolutional neural network, was trained with 2763 melt pool images captured from a laser melting powder fusion build using a serpentine scan strategy. The model is validated through 2926 new images collected from a different part in the same build using ‘island’ serpentine strategy with predictive accuracy of 91%. Compared to a traditional image analysis method, the processing time of the validation images is reduced by 90 %, from 9.72 s to 0.99 s, which gives the feedback control a reaction time window of 0.34 ms/image. Results show the feasibility of the proposed method for a real-time closed loop control of L-PBF process.

Advanced sensors are increasingly used in aeronautical facility. Broader acquisition of images improves the quality and quantity of remote sensing images and makes a large number of fine remote sensing images available. Although semantic segmentation models based on deep learning has been a great success in computer vision, there are still a lot of difficulties compared to regular visual senses. In this paper, we propose a novel end-to-end model for remote sensing image segmentation of building footprints and a loss function based on IoU to improve the performance. Semantic labels for building structures are being accurately allocated by our method. We make control experiments on Map Challenge dataset with other semantic segmentation model based on recent deep learning advances. Our proposed method generates 94.73% of IoU on the Map Challenge Dataset's validation set, which outperforms all previous models in literature.

In this paper, the off-state avalanche stability is investigated and a time-dependent avalanche instability is observed at −40°C in SOI lateral IGBT. Combining results from measurements with those from TCAD simulations, the mechanism for avalanche instability is revealed. Two optimization schemes are proposed to suppress or eliminate the unstable avalanche phenomenon. The unstable avalanche can be suppressed by employing a thick buried oxide layer and V CE shift can be reduced from 54V to 11V. Another scheme is that replacing P-type substrate by N-type substrate, and then V CE shift can be completely eliminated.

The type II short-circuit failure mechanism of trench gate insulated gate bipolar transistor (IGBT) is investigated in detail. By using physical TCAD simulations and experiments, it shows that high hole current goes along the trench gate which restructures the electric field distribution under the bottom of the trench gate. As a consequence, a dynamic peak voltage appears under the bottom of trench gate, which results in the gate oxide breakdown during the type II short-circuit process. An optimization structure with reducing the trench gate depth is proposed to suppress the dynamic peak voltage and improve the type II short-circuit robustness for trench gate IGBT.

In consideration of the convenience of drivers’ operations, a velocity guide driving model for collision avoidance in non-signalized intersection is proposed. Based on reduplicate game theory, a profit function consists of safety, rapidity, controlling indicators is redefined. Pareto optimality is adopted to obtain the strategies which maximize the profit for the whole game. Simulation results show that the model is effective and gives a more comfortable vehicle-cross process.

In this paper, a novel liquid metal current limiter (LMCL) is proposed allowing for the large rated current and considerable fault current energy dissipation in DC power system. According to this topology, the LMCL consists of a fast switch and several liquid metal current limiting units in parallel. Firstly, a simulation model of this LMCL is built to calculate the fault current commutation process and it is then experimentally verified. Based on this model the duration of current commutation is analyzed allowing for different factors including rising rate of current, contact types of main switch and the inductance of branch circuit, which directly affects the design of parallel current limiting unit to avoid its early arc ignition before the completion of current commutation. Finally, Its current limiting performance is discussed with different number of channels in LMCL unit, it indicates that arc voltage increases with the channels quantity of LMCL unit.

The influence of forward current freewheeling time on the reverse recovery di/dt robustness of Superjunction (SJ) MOSFET body diode is investigated in detail. It is found that the maximum di/dt capability of body diode is improved dramatically with reducing the forward current freewheeling time. To explore this phenomenon, physical TCAD simulations and experiments have been carried out. It shows that there is an extremely inhomogeneous reverse recovery behavior between the cell region and the termination region in the SJ-MOSFET, which results in a destructive dynamic avalanche behavior at the edge of the termination. The reduction of the forward current freewheeling time can suppress the inhomogeneous phenomenon and impact the reverse recovery di/dt robustness of body diode. Thereby, it points out that the real di/dt robustness of SJ-MOSFET body diode should be evaluated with the long enough forward current freewheeling time.

This paper reports on implementation of sinusoidal battery charging technique for Wireless Power Transfer (WPT) with maximum efficiency control algorithm. In sinusoidal current charging technique, the battery current varies continuously depending on charging frequency and current. Therefore, to improve the overall efficiency of the system, load impedance has to be matched with source impedance continuously depending on charging current frequency. In this paper, an algorithm is presented to maximize the efficiency of the system. For concept A, algorithm varies both frequency and phase shift of the inverter output voltage to minimize input power, whereas output power is kept constant by a dc-dc converter. For concept B, the battery current is controlled by switching frequency, and the algorithm optimizes the phase shift of inverter to improve efficiency. To verify the proposed system, a brief mathematical analysis is provided. Further, a laboratory prototype of a 600W power rating is developed, and results are presented to verify the effectiveness of the concept.

During the cycling of lithium-ion batteries, the internal resistance increases due to the deposition of Solid Electrolyte Interface (SEI), which causes the power fade of batteries. In this paper, lithium ion batteries with different levels of power fade are compared, their internal resistance is measured and used to predict their thermal behavior. Based on numerical methods, temperature rise and heat generation of lithium ion batteries are presented under constant C-rate discharge and power pulse. Furthermore, the calculated results are validated through measurement. Finally, it can be concluded that higher level of power fade leads to higher heat loss and lower output power and an optimal DOD range is advised for the purpose of reducing temperature rise.

This work provides a physical insight into the failure of Superjunction MOSFET (SJ-MOSFET) for full bridge inverter system during an reverse recovery behavior. The inspection of failed SJ-MOSFET reveals burnt-out points in the vicinity of the device edge termination. To explore this result, physical TCAD simulations have been carried out. It is found that there is an extremely inhomogeneous reverse recovery behavior between the active region and the termination region in the SJ-MOSFET. And a destructive dynamic avalanche phenomenon is occurred at the edge of the termination, which is identified as responsible for the observed failure. Finally, an optimization termination structure with P+ dynamical field limiting ring (DFLR) is proposed to suppress the dynamical avalanche and the reverse recovery robustness of the proposed structure is improved by 200%.

A composite device structure on Silicon-On-Insulator (SOI) layer named Self-adjust Conductivity Modulation SOI-LIGBT (SCM-LIGBT) is proposed. It can be divided into three parts: the normal LIGBT region (NLT structure), the EM-NMOS region (ENM structure) and the diode region (DIO structure). The drain of the ENM structure is connected with the n+ emitter of the NLT structure while the p+ region in the emitter side of the NLT structure is connected to the anode of the DIO structure. The gates of the NLT structure and the ENM structure are connected together and they acted as the gate of the proposed SCM-LIGBT structure. In the on-state, the NPN parasitic bipolar structure of the NLT structure is triggered and the conductivity modulation is dramatically enhanced, which leads to the reduction on the on-state voltage. In addition, due to the base voltage of the NPN parasitic bipolar structure in the proposed device can be clamped at the forward threshold of the series diodes, therefore the latch-up issues can be immunized to guarantee the forward-biased safe-operating-area (FBSOA). The experiments demonstrate that the proposed SCM-LIGBT achieves the Vf lower than 1.18V at J A =150A/cm2.