Search

Word embedding representation has been a research hot spot in the field of computer vision and natural language processing. The semantic quality of representation severely affects the effects of machine translation and image caption. The key factors affecting the quality of word embedding representation are capture of contextual semantics and solution of the problem of polysemy. In this paper, the principles of the current mainstream word embedding representation methods were explored, such as sparse representation (One-Hot), static representation (Word2Vec, GloVe), dynamic representation (ELMo, BERT) representation and based on position information (NN.Embedding-Position Embedding). The impact of those word embedding representations on image caption task was researched attentively. And those representations methods were compared experimentally using the MSCOCO data based on the same image caption model. The pros and cons of these methods were judged according to the accuracy and richness of the extracted label data. Experimental results show that the BERT model representation with dynamic bidirectional semantics has the best performance in image caption tasks, but it requires longer training time. The NN.Embedding representation has the second best image semantic description effect, and its training speed is the fastest.

Computerized systems and software, which allow optimizing and planning the processes of production, storage, transportation, sale, and distribution of goods, have emerged in the industry. Scheduling systems, in particular, are designed to control and optimize the manufacturing process. This tool can have a significant effect on the productivity of the industry because it reduces the time and cost through well-defined optimization algorithms. Recently, the applicability of blockchain technology has been demonstrated in scheduling systems to add decentralization, traceability, au-ditability, and verifiability of the immutable information that this technology provides. This is a novel contribution that provides scheduling systems with an additional layer of security. With the latest version of Hyperledger Fabric, the appropriate levels of permission and policies for access to information can be established with significant levels of privacy and security, which prevent malicious actors from trying to cheat or abuse the system. Different alternatives exist to manage all processes associated with the operation of a blockchain network, and among them, providers of blockchain as a service have emerged. Chainstack stands out for its simplicity and scalability features to deploy and operate a blockchain network. Our goal in this work is to create a solution for secure storage of and access to task-scheduling scheme on the consortium blockchain and inter-planetary file system as a proof of concept to demonstrate its potential and usability.

In recent years, synergy theory has made great progress in the field of power electronics, and its good robustness provides a new idea for motor control research. In this paper, we take use of rapid convergence of manifold, and introduce it into the control of the speed and flux linkage of the traditional deadbeat control (DC stands for deadbeat control and IDC stands for improved deadbeat control), aiming to eliminate the overshoot problem in the traditional DC. Based on the above work, this paper deduces the improved control algorithm based on synergy theory for deadbeat prediction. The simulink simulation platform is built. The feasibility of the theory is verified by algorithm and simulation analysis.

In order to improve the heat management level and monitoring quality of urban heat network, a genetic algorithm based PID hot network monitoring system is designed. The system uses PLC as the controller and uses the analog input module to collect the outdoor temperature in real time. According to the average temperature curve of supply and return of the secondary pipe network, the genetic algorithm is used to optimize the PID parameters online and control the average temperature of the secondary pipe network supply and return water. The simulation results show that the PID based on genetic algorithm has better overshooting, faster control time and stronger robustness than traditional PID control, and has good application value.

Under the background of engineering education professional certification, this article puts forward the construction of new training model for automation professionals. This model aims at cultivating engineering innovation and entrepreneurial talents. Through a series of measures such as training program optimization, curriculum system renewal, teaching method reform and practical teaching enhancement, it has realized the cultivation of professional talents with high quality, strong ability and quick adaptability. The practice shows that the new talent training model can effectively improve students' comprehensive quality, practical and innovative abilities, and employment competitiveness.

Mihcak et al. proposed a locally adaptive window-based denoising method using maxsimum likelihood (LAWML) with low complexity. However, LAWML was based on decimated wavelet transform (DWT) without taking account of the interscale dependencies. In this paper, we propose a variant of LAWML, namely MPLAWML, based on multiscale products. We improve LAWML by extending DWT to undecimated wavelet transform (UWT) and multiplying the adjacent wavelet subbands to exploit the wavelet interscale dependencies. In the multiscale products, edges are enhanced and noise is weakened. Thereafter, a product threshold is calculated for each product subband and is used on the products coefficients to identify significant features. Then LAWML is applied to process those wavelet coefficients, which are greater than the corresponding products thresholds. Experiments show that the proposed algorithm has more robustness to noise, achieves better visual effects than LAWML and has competitive performance compared with the state-of-the-art wavelet-based denoising algorithms.

A CPW fed ultra wideband (UWB) antenna which is working on the bandwidth 0.2GHz–3GHz with two narrow-band suppression is presented in this paper, and there are some analysis using characteristic mode theory. The antenna is a compact printed monopole antenna, and it produces narrowband suppression through the embedded split-ring resonator and U shape slot in the radiation patch. The simulation results show that, in addition to the specific stop-band frequency, antenna voltage standing wave ratio (VSWR) is less than 2∶1 within the scope of the whole working bandwidth. Therefore, the antenna designed in this paper has good application prospect.

In the paper, the planar parasitic coupling multi-resonator configuration yielding broad BW (Bandwidth) for wireless communication system is presented. The proposed antenna, a rectangular driven element and the same two semicircular parasitic elements with a pair of arc-shaped slots, fed by a coaxial probe, is etched on a thin circular dielectric substrate Teflon. In order to enhance bandwidth of micro-strip antenna, multi-resonator technique is employed. In this case, we introduce two identical semicircular parasitic elements which are placed at the same distance from the two non-radiating edges of the rectangular driven element. Besides, cutting the same two arc-shaped slots in the two parasitic elements will be conducive to excite adjacent modes of the fundamental resonant mode for the proposed antenna. Results show that more modes close to the fundamental resonant mode will appear. For the proposed antenna, impedance bandwidth over 10% of the Operating frequency is obtained. Compared to conventional antennas, the antenna has better performance in the requirement of bandwidth and is simple in configuration. Moreover, the peak gain and 3-dB beam-width of the proposed antenna are approximately 6.5dBi and 90°, respectively.

Based on Simplifying the antenna modes through the application of characteristic modes approach, we propose a novel method which synthesize the Gauss filter to design antenna equivalent circuit. We take the dipole antenna as an example to expound the method. Firstly, from the analysis of the characteristic modes approach, mode 1 and mode 3 are found to be the main radiation modes for a dipole antenna. And then, band pass filter equivalent circuits are designed in the light of the modal coefficient graph of each mode. Finally the two band pass filter equivalent circuits of mode 1 and mode 3 can be combined in parallel to be the equivalent circuit of the dipole antenna. The simulation results of the equivalent circuit are quite consistent with the original antenna for the bandwidth and impedance.

We proposed a robust algorithm for the phase unwrapping of interferometric SAR data for borehole application using Bayesian Markov Random Field (MRF) models. Our contribution to the existing algorithm consists on the exploitation of the Borehole InSAR coherence map in order to generate a more adaptive mask for each decomposition level. Moreover, we presented a general probabilistic framework to phase unwrapping problem. Bayesian estimator was applied to recover the desired phase, as the optimal field solution of the maximum a posteriori (MAP) estimation criterion. An iterative method that minimizes a general energy function is proposed.

In this paper the main factors that affect our revenue growth from both qualitative and quantitative point of view. In the quantitative analysis, we using of EVIEWS6.0 software, collecting 1990–2009 data, to empirical analysis the economic factors (mainly including gross domestic product, the financial expenditure and the retail price index) which affect our revenue growth, has been gotten the multiple linear regression model of the impacts of China's tax revenue growth. And then we use the model to validate the 2010 tax income of our country.

Nowadays the mobile equipment has become an essential tool of providing access to business resources from the organization's internal network and other services. Hence the security of wireless communication becomes more and more important. This paper demonstrates how to implement a mobile access system based on multi-factor authentication for security mobile communication,at the same time, improving IKE process to make the system more efficient.

Based on the analysis of coal and electricity industry peculiarities and present conflicts, according to the Stackelberg game features demonstrated by coal and electricity supply chain profit distribution of joint venture, this paper uses bi-level programming theory to construct the coal and electricity supply chain profit distribution model, and gives the chaotic searching algorithm, and then uses the simulation example test the rationality of the model and put forward suggestions for improvement.

In this paper, the problem of attitude control of spacecraft with the unknown inertia matrix is studied. The model of kinetics in which the reaction wheels as executive component of spacecraft attitude control is considered. Based on the Lyapunov stability theory, the law of controller and estimation values of inertia matrix is obtained. Numerical simulation results show the validity and effectiveness of this method. To demonstrate the robustness against the effective of bounded noise of the proposed technique, it is applied to the model system and simulation results are obtained.

Satellite telemetry parameters are the basis of judging whether satellites are in order or not. And parameters' trend prediction is the precondition to predict the condition of the satellites. Aiming at the need of getting data trend in the project, the comparability of parameters and metabolism was put forward to improve precision of prediction, as well as to build a new prediction model which was adopted to calculate the trend. Through resembling using the data from project practice, new idea is valid and dependable.

Both kinematics and measurement model help process the trajectory data and estimate aerodynamic parameters of vehicle. In this paper, three dynamics are proposed, which are aerodynamics parameter's model, acceleration's model and α - β - γ model. The measurement model is TDRS (Tracking and Data Relay Satellite), which produces a complicated trajectory in reentering purpose. A fuzzy IMM algorithm is proposed in this paper, which through a fuzzy inference to get the matching degree of every UKF filtering model, instead of calculating model transition probabilities in IMM. With the matched degrees, the estimation from each filtering is weighted to get maneuvering target overall estimation and its covariance. Simulation results show that the FIMM algorithm performs well in tracking re-entering vehicle and estimating the aerodynamics parameters as well.

In this paper, a tracking system with multi-sensors is presented, in which a sub-orbit USV (unmanned space vehicle) of wave-rider shape is tracked by a TDRS (Tracking and Data Relay Satellite) and ground stations. Because of high lift-drag ratio and maneuverability, the vehicle, once is used in reentering purpose, a complicated trajectory will be produced and cause big challenges for tracking facilities. Although TDRS has more coverage capability than ground stations, unlike tracking stations fixed on the ground, when its antenna tracks target, the platform it located on will do some attitude maneuver to balance the effects caused by the antenna's moving. This will lead some errors in tracking the target and should be taken into account when processing the data. To solve these problems, a fusion strategy of improved IMM (Interacting Multiple Model) with different kinematics and measurement models is designed in this paper, which helps to process the trajectory data and estimate aerodynamic parameters of the vehicle. In the dynamic model, both trajectory parameters and aerodynamics parameters are calculated, the measurement models are either TDRS or ground tracking stations. Meanwhile, in the adaptive IMM algorithm, an adaptive method for calculating transition probabilities of time-varying transition model is designed. Also the Iterated Sigma Point Kalman Filter (ISPKF) is used helping make the system more robust and perform better. Simulation results show that the proposed algorithm performs well in tracking high lift-drag ration re-entering vehicle and estimating the aerodynamics parameters as well.

Based on the analysis of the importance of project termination decision and the review of related research, this paper constructs an R&D project termination decision model based on Kohonen neural network consisting of the principles of a system of evaluation indexes, the quantification of eigenvalues of the evaluating indexes, the basic principle of Kohonen and the establishment and application of the model. Then the paper constructs the model by a case example based on the establishment and application of the model to confirm the validity of the model and draws a conclusion that the method it used was more effective compared with other methods.

The construction of Smart Grid will comprehensively enhance the intelligent level of every step in the power grid of our country. The data forecasting ability determines the quality of smart grid. As a result of this, this paper constructs the background value of grey GM (1, 1) model by using threepoint Gaussian quadrature formula, improves its accuracy and the data quality. It indicates that reconstruction of the background value is the key factor affecting prediction accuracy and applicability.

Feature extraction for object representation plays an important role in image denoising system. As the spatial histograms consist of marginal distribution of image over local patches, object texture and shape are simultaneously preserved by the spatial histogram representation. In this paper, we propose methods of compressive sensing for spatial histogram-based object detection. We employ Fisher criterion to measure the discriminability of each spatial histogram feature and calculate features correlation using mutual information. In order to construct compact feature sets for efficient classification, we propose informative selection algorithm to select uncorrelated and discriminative spatial histogram features. The proposed approaches are tested on two different kinds of images. The experimental results show that the proposed approaches are efficient in feature extraction and image denoising.

Traditional methods for Natural Image Understanding can both be computationally expensive and lack robustness. A recently proposed technique for Natural Image Understanding, based on sparse coding, is computationally less expensive and has demonstrated the capability to correctly identify objects from particular types of noisy images. In this paper we examine the ability of this sparse coding technique to handle broader challenges that are likely to be relevant for Natural Image Understanding systems in practice. We find that it remains robust for varied viewing angles, expressions, and illumination. However, identification accuracy suffers when the size of the training database is significantly less than the size of the testing set. We propose a simple technique that could improve the reliability and accuracy of sparse coding based Natural Image Understanding systems.

In this paper, the application of Independent Component Analysis (ICA) for video frame sequence processing is investigated and a novel scheme based on ICA for detecting and locating text in video is proposed. The author analyses the advantage of novel scheme over current scheme, the results of simulation experiment illustrate that analysis is correct and novel scheme is more effective than current scheme.

Malware has become the centerpiece of security threats on the e-commercial business. The focus of malware research is shifting from using signature patterns to identifying the malicious behavior patterns. Many researcher extract behavior pattern from system call sequences to identify malware from benign programs with data mining techniques. Most system call tracing tools must run alongside the malware in the same system environment and could be easily detected by malware. In this paper, we propose a new system calls tracing system based on the full virtualization via Intel-VT technology. Malicious samples are running in a GuestOS and they can not detect the existence of system call tracing tool running in the HostOS. We collect a system call trace data set from 1226 malicious and 587 benign executables. An experiment based on the SVM model shows that the proposed method can detect malware with strong resilience and high accuracy.

Copulas are extremely versatile, and can be used as an analytical tool in a broad range of financial situations such as risk estimation, credit modeling, pricing derivatives, and portfolio management. The literature on copula is mostly devoted to probabilistic theory, to inference methods or to applications in financial risk management. The thesis presents a combination of these three aspects.

In order to make aircraft track ideal flight path precisely when approaching and landing on carrier, automatic carrier landing system (ACLS) should have adequate transient response and the ability of restraining gust disturbance. So active disturbance rejection controller (ADRC) was used to design longitudinal automatic carrier landing guidance law, and a simulation model of ACLS was developed in the Matlab/Simulink environment. The particle swarm optimizer (PSO) algorithm was then used to determine a set of optimal gains for the guidance law. Simulation results show that the ACLS using ADRC guidance law has good characteristics. The performance of step response and the ability of restraining gust disturbance are better than the PID guidance law.

One of the major threats to cyber security is distributed denial of service (DDoS) attacks. In this paper, we propose a new algorithm based on the persistent increment tendency of DDoS traffic. Our scheme can detect a DDoS attack in its early stages when the attacking packet's attribute value has no distinct features. It can differentiate DDoS from flash crowd traffic. This scheme detects DDoS attacks with on-line and distributed characteristics. Simulation shows the algorithm's validity and accuracy.