Search

We propose Iterative Homography Network, namely IHN, a new deep homography estimation architecture. Different from previous works that achieve iterative refinement by network cascading or untrainable IC-LK iterator, the iterator of IHN has tied weights and is completely trainable. IHN achieves state-of-the-art accuracy on several datasets including challenging scenes. We propose 2 versions of IHN: (1) IHN for static scenes, (2) IHN-mov for dynamic scenes with moving objects. Both versions can be arranged in 1-scale for efficiency or 2-scale for accuracy. We show that the basic 1-scale IHN already outperforms most of the existing methods. On a variety of datasets, the 2-scale IHN outperforms all competitors by a large gap. We introduce IHN-mov by producing an inlier mask to further improve the estimation accuracy of moving-objects scenes. We experimentally show that the iterative framework of IHN can achieve 95% error reduction while considerably saving network parameters. When processing sequential image pairs, IHN can achieve 32.7 fps, which is about 8x the speed of IC-LK iterator. Source code is available at https://github.com/imdumpl78/IHN., Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR 2022)

The sensor-based human activity recognition has been wildly applied in behavior tracking, health monitoring, indoor localization etc. Using activity continuity to assist activity recognition is an important research issue, in which the activity transition matrix which describes the activity transformation in real scenarios is the most important parameter. Aiming at the problem that the current classic transition matrix learning algorithm cannot fuse weights of sample classification results, a weighted transition matrix learning algorithm is proposed in this paper. First, the basic definitions of an improved Hidden Markov Model (HMM) which fuses weights of classification results are given. Then, the recursive formula of transition matrix learning is derived, and the learning algorithm W-Trans is put forward. Finally, the proposed algorithm is simulated with the public data sets. The evaluation results show that the proposed algorithm outperforms the classical Baum-Welch algorithm under evaluation metrics of both the cosine similarity and the euler distance. By applying W-Trans to current activity recognition post-process methods, the advantage of our method is verified.

Reversible watermarking is one of the main methods to authenticate and secure digital media security and revert to its original state without any losses. There are multiple reversible watermarking schemes proposed over the years. Often the proposed reversible watermarking scheme was focused on general grey-scale images. Their performance on medical images is not often evaluated. This study compares the performance of various general reversible watermarking schemes using multi-frame medical images and general grey-scale images in terms of 3 main factors, (1) computational complexity, (2) imperceptibility, and (3) embedding capacity. The result has shown that the Least Significant bits have the highest imperceptibility and embedding capacity. However, it is the least secure where outside can modify the watermarked images as long as they have the essential information.

In the mordem society, physical infrastructure and information technology are inseparable. The concept of cyber-physical systems (CPSs) is then proposed and has been widely concerned by researchers in recent years. Various models have been introduced to meet the actual needs. In one type of those models, the physical part and the cyber part are coupled and influence each other. The cyber part monitors and controls the physical part, but at the same time it may also bring certain harm; the fault in one part may also be transmitted to the other and affect the performance of the counterpart. Here, this paper introduces an asymmetric interdependent model and studies how the coupling pattern of CPSs affects the system robustness. In addition, the coupling pattern of CPSs is optimized by using the simulated annealing (SA) algorithm to reduce the performance loss. The results of this paper may find applications in the future CPS planning.

Recently, a cybertwin based cloud-centric network architecture for future generation networks was proposed, which can work in a distributed way through a realtime multi-agent trading platform to allocate hierarchical resource, such as computing, caching, communications resource. The current network operator centralized resource management mechanism, however, cannot sufficiently manage the resource in a flexible manner. In this work, we propose a barter and combinatorial auction based hierarchical resource trading mechanism for cybertwin network. A unified bid language is further presented to express various resource requests and supplies. We formulate the total payoff maximization problem as an integer programming problem. To solve the problem within a polynomial time, we transform it into a minimum cost multi-commodity flow and solve it efficiently. Simulation results show that the proposed model can obtain the optimal solution real-time.

To develop a high-efficiency synchronous reluctance motor (SynRM) drive system, a novel maximum power factor control (MPFC), which is based on current angle control with stator flux and stator resistance estimators, is proposed in this study. First, a conventional maximum power factor control (CMPFC) system using a saliency ratio of the SynRM to generate a fixed current angle command is introduced. Since the saliency ratio needs offline pre-preparation and can’t be adjusted automatically, it is difficult to improve the performance of the MPFC in different operating region owing to the increasing of manufacturing cost and time-consuming. Therefore, an online maximum power factor searching control (MPFSC) based on current angle control is designed for the speed control of a SynRM. Moreover, a proportional-integral (PI) speed controller is adopted to generate the stator current command, and the proposed online MPFSC is employed to produce the current angle command. Finally, from the experimental results, the current angles of the optimal power factor (PF) can be effectively obtained online at different speed and load torque conditions.

Under the variety of information and communication technologies, the control center can securely and reliably operate power grid during the system disturbances and natural events. Specifically, when the initial failures are detected by cyber monitoring, the system will timely take emergency protecting operations to restrain the spreading of failures. Although cyber network makes the system more robust, they also increase the risk from the cyber network. In this paper, we firstly consider a control strategy into the cascading failures model. Then, we study how the cyber attack — False negatives attack confuses the control center, and makes the control center can not take emergency operation timely, causing the failure continue spreading. We propose an optimization problem to model the above assumptions and, by solving the optimization problem, study the impact of different attack scenarios on power system.

Ocean reverberation is an important factor to affect the detection performance of active sonar, especially, in ice-covered sea areas. Because the interaction between ice and water will produce strong reverberation field, which have a great impact on the acoustic detection and communication performance. Physical acoustic method is firstly used to study the reverberation characteristics under the rough ice surface in this paper. Based on Kichhoff approximation, a Helmholtz scattering model is established under rough ice surface, and obtain the reverberation signal under ice, and analyzed its statistical and related characteristics. In order to simulate the ice surface which is in a fluctuating state, the stochastic process description method is used to describe the random fluctuating rough surface According to the reflection theory of ice-water interface, obtaining the relationship between the reflection coefficient of ice-water interface element and the incident angle function. The Hotz integral acoustic scattering model is used to obtains the reverberation signal under ice and analyze its correlation characteristics. At the same time, the correlation characteristics of the reverberation signal under ice is analyzed by analyzing the test data in Songhua River. The analyzed results show that the time correlation radius decreases and the correlation decreases with the increase of the signal frequency, and the spatial correlation coefficient shows oscillation attenuation relationship with the increase of the array spacing.

Binary neural networks can effectively reduce the number of required parameters but might decrease the classification accuracy. To solve the problem, we propose a dual-path binary neural network (DPBNN) in this paper. Experimental results show that our DPBNN can outperform other traditional binary neural network in CIFAR-10 and SVHN dataset. The proposed network is simple, so it is suitable to be implemented on embedded systems or SoC designs.

Micromachined gyroscope is the core component of modern guided weapons, but the launching process of guided weapons is accompanied by huge acceleration overload. Aiming at the problem that the moving mass of micromechanical gyroscope is greatly affected by overload under the condition of large overload, a MEMS cup-type vibrating gyroscope structure with anti-high overload is designed. Combined with the principle of four-abdominal motion, the working principle, vibration characteristics and anti-high overload characteristics of cup-type gyroscope structure are analyzed. The finite element model of Silicon microcup gyroscope is established in ANSYS finite element analysis software. Modal analysis and harmonic response analysis are performed respectively. The simulation analysis results show that the natural frequency difference of the Silicon microcup gyroscope between driving modal and sensitive modal is 0.8kHz, the frequency matching of working modal is good. The impact response of this structure under the action of a half-cycle sinusoidal acceleration impact load is analyzed according to the principle of impact dynamics, the maximum stress and maximum displacement of the resonant structure under the transient impact of 100000g are 11.38MPa and 8.06nm respectively. It is proved that the structure has excellent impact resistance.

In the past decade, the rapid popularization of smartphone has provided a promising direction for human activity recognition. Despite identifying a variety of movements without any complicated wearable device, the smartphone-based activity recognition is still deeply affected by the differences between users and phone locations. To overcome this problem, post-process attempts to correct the errors in the classified activity sequence. In consideration of both the activity sequence continuity and the recognition result confidence, we propose WOODY, a novel post-process method that locates and corrects the errors in a classified activity sequence just like Woody Woodpecker pecking holes to catch the pests. In our method, the recognition result is considered as the weighted observation state, and a weighted observation hidden Markov model (WOHMM) is built to model the classified activity sequence. Consequently, a sequence labeling algorithm of the WOHMM is also designed to modify those recognition results with low confidence. To validate the effectiveness of WOODY, we make a series of contrast experiments on two public data sets collected from real scenarios. The results show that WOODY is not only able to improve the recognition accuracy but also significantly enhance the robustness.

Spiking neural networks (SNNs) are artificial neural network models that are closely mimic natural neural networks. LIF (Leaky Integrate-and-fire) neuron model, population coding and Tempotron supervised learning rules are used to construct a spiking neural network for visual color feature classification based on RGB-HSV (Red, Green, Blue -Hue, Saturation, Value) model. The product of a momentum learning rate and the last weight change is proposed to speed up the training of the SNN. Test results based on a common data set show that the accuracy of the SNN can be up to 90%.

Cascaded H-bridge topology has been used in grid-tied converter for battery energy storage system due to its modular structure. To fully utilize the converter’s modularity, we propose a hierarchical distributed control architecture that consists of primary control and secondary control. Primary control ensures correct current tracking, but may lead to unequal power output among modules. A distributed secondary control based on consensus algorithm is presented to establish equal power sharing among modules. Further, a carrier phase shift control is implemented to achieve multilevel output voltage and harmonic reduction. Finally, the effectiveness of the presented control strategy is verified through experiments.

The large random errors in a typical MEMS gyros under the ranges of relative humidity condition. In order to solve the large random error under the ranges of relative humidity, a method based on adaptive threshold denoising and neural network is proposed in this paper. Full relative humidity experiment is carried to test LSM6DS3 MEMS gyros. Then the adaptive threshold denoising is used to separate the drift error and drift data modeled with RBF neural network. The experimental results show the effectiveness of the method, and it is of great significance in improving the performance of MEMS gyros.

In this work, we propose an automatic radio map construction system based on crowdsourcing Pedestrian Dead Reckoning (PDR) traces, which does not rely on priori knowledge of floor plans and is robust to inaccurate PDR traces. In this system, we propose to process some opportunistic PDR traces, in which users walk through the building, to generate parts of road paths by translating, rotating and scaling the traces based on the opportunistic GPS locations and gate points as landmarks. Then, we further extend the coverage of road paths by processing the PDR traces entirely obtained indoor by compensating the turning errors and merging the PDR traces based on the similarity of WiFi fingerprints. With such procedures, we can accurately generate indoor road paths of a large-scale building and construct the radio map based on these road paths. Our proposed method achieves a median accuracy of 2.8m and mean accuracy of 2.9m for the constructed road paths. By fusing GPS, PDR, and WiFi fingerprinting with the crowdsourcing radio map, we achieve a median positioning accuracy of 2.9m and mean accuracy of 3.4m without site surveying, which significantly outperforms the positioning algorithm by merely fusing GPS and PDR.

A long-primary single-sided linear induction motor (SLIM) for a high temperature superconducting (HTS) magnetic suspension system is analyzed by numerical method and finite element method (FEM). An optimal design of SLIM is considered, in which the structures and dimensions of the secondary have been invistugated.

The robust power allocation problem is often solved by different optimization approaches under a convex optimization model. But in this study, we use a distributed projected dynamic system (PDS) to describe this model and realize power allocation by designing a stable controller using Lyapunov function and linear matrix inequality (LMI) for the PDS. The controller can follow our defined target SINR and keeps the quality of service (QoS) required by primary user (PU) under the channel and the interference uncertainties (including feedback error). The simulation results illustrate that our proposed controller can realize power allocation with better performance compared with iterative waterfilling algorithm (IWFA).

Time-reversal communication is considered as a potential Green wireless communications scheme for cognitive radio networks. The network base station utilizes time-reversal communication to exploit multi-path propagation, in order to provide spatial focusing at an intended cognitive radio. This reduces interference to other radios in the network. However, time-reversal spatial focusing performance is dependent on robust channel estimation. Under time-varying channel conditions or imperfect channel estimation, the performance of time-reversal communication deteriorates immensely. To ameliorate this deterioration, we design a non-stationary time-varying non-causal Wiener filter based on the time-varying spectrum. The time-varying spectrum is obtained by first modeling the time-varying channel as a locally stationary process. Which means that over small time intervals the channel is approximately stationary, and correlated inside these stationary intervals. Consequently, the time-varying spectrum can be easily calculated by estimating the covariance of the Wigner-Ville distribution of each locally stationary process. Based on that premise, the goal of this paper is to investigate through simulation, the performance of the proposed Wiener filter versus the conventional optimal Wiener filter when the time-varying channel is modeled as a locally stationary process. The performance is derived by computing the symbol error rate (SER), minimum mean square error (MMSE) and the output versus input signal-to-noise ratio (SNR).

In industry, evaluating candidate architectures for automotive embedded systems is routinely done during the design process. Today's engineers, however, are limited in the number of candidates that they are able to evaluate in order to find the optimal architectures. This limitation results from the difficulty in defining the candidates as it is a mostly manual process. In this work, we propose a way to synthesize multilevel, multi-perspective candidate architectures and to explore them across the different layers and perspectives. Using a reference model similar to the EAST-ADL domain model but with a focus on early design, we explore the candidate architectures for two case studies: an automotive power window system and the central door locking system. Further, we provide a comprehensive set of questions, based on the different layers and perspectives, that engineers can ask to synthesize only the candidates relevant to their task at hand. Finally, using the modeling language Clafer, which is supported by automated backend reasoners, we show that it is possible to synthesize and explore optimal candidate architectures for two highly configurable automotive subsystems.

the safety assessment is of great significance for both design progress and airworthiness certification of civil aircraft. Given this, the system safety analysis based on Fault Tree Analysis (FTA) and Markov Analysis (MA) is carried out on the Full Authority Digital Engine Control (FADEC) system of an aircraft engine. With the help of safety analysis software Isograph workbench, the FTA and MA methods are used to analyze a disastrous failure state "control system failure" of a FADEC system. By using FTA method, the failure is analyzed further both qualitatively and quantitatively; the causes of the disastrous failure are identified, and the failure rate of top event is generated. The established Markov model can obtain the probabilities of the occurrence of different system states and the probabilities of failure states of the system under different maintenance intervals. Finally, we can make reasonable recommendations for maintenance policy.

In order to solve the inconsistency of the battery pack in the traditional battery energy storage system, a new type of battery module energy storage system topology and control strategy based on flexible grouping is proposed——Modular Battery Energy Storage System Based on One integrated Primary multi-secondaries transformer. Different from commonly used full-power control mode, the part-power control mode is adopted in this paper. The current flowing through the switches is the difference of each battery module current and the average current of all battery modules. By applying part-power control method, the system volume, cost and loss will be effectively reduced. Based on the dual active bridge converter and one integrated primary multi-secondaries transformer, phase shifting control strategy is implemented to control part of the charging and discharging current of each battery module independently, which makes each battery module work under the best condition and helps in improving the energy utilization. Based on analyses of the data of the passenger car battery after 3 years of actual operation, Simulation results show the correctness and superiority of this topology and its control method.

In this paper, we study spectrum overlay access to share spectrum between primary users (PUs) and secondary users (SUs). Our goal is to maximize the average throughput by optimal power allocation within finite time duration. To do this, we formulate this optimization problem as a Markov Decision Process with continuous state. An approximate value method with pre-allocation mechanism is proposed, which can effectively protect the PUs by obtaining a continuous closed-form solution rather than a discrete one. Numerical results show that the proposed algorithm exhibits better performance than traditional methods while guaranteeing non-interference to PUs.

Steganography has various useful applications. In this paper, a steganographic algorithm with high payload and low change rate is proposed. Three pixels are processed as a unit, the exclusive-or result of the second least significant bits (SLSBs) of the pixels in that unit is regarded as the flag. According to the value of the flag, as well as the Hamming distance between the least significant bits (LSBs) and the secret messages, three bits of secret message are embedded. Secret message can be easily extracted depending on the value of flag. Local texture is also considered to achieve adaptive embedding. The advantage of the proposed algorithm is that at most one pixel will be modified when embedding three bits of message into three pixels. Simulation results showed that the proposed algorithm decreases the probability of detection for steganalysis methods and the security has been much improved.