Search

Visible infrared person re-identification (VI-ReID) is an extremely challenging task. VI-ReID suffers from two challenges. One is the cross-modality discrepancy due to different camera spectrums, the other is the intra-modality variation caused by the noise of background clutter or occlusion. We propose a global-level and local-level constraints network (GLoC-Net) to learn discriminative feature representations. It mainly contains two aspects. 1) We employ a non-local attention mechanism for extracting shared features to mitigate the cross-modality discrepancy, and present the division operation of local features to alleviate the problem that the non-local attention mechanism is less robust to noise. 2) We propose joint constraints of global-level and local-level to alleviate the intra-modality variation, which makes the algorithm more robust to noise. Experiments demonstrate that the superior performance of proposed method compared with the state-of-the-arts.

Improving the development and utilization of renewable energy is one of the critical goals in the green industrial revolution of China. The energy efficiency measurement and the detection of influence factors are the very important issues to accomplish this task. In this paper, feature recognition is applied to determine the factors that affect the energy efficiency. Based on the feature recognition algorithm, data mining methods are used to derive knowledge from a dataset of 24 provinces or cities in China. In the process of data mining, the following three problems are addressed: 1) the optimal feature subset is selected from the original feature set that affects the energy efficiency; 2) the energy efficiency based on the optimal feature subset is evaluated; and 3) the energy efficiency in China is predicted by seven good-fitted classification models, whose accuracy rates are higher than 90%. Combining the results of feature selection and energy efficiency prediction, the strategy and policy to improve the energy efficiency in China is suggested. Green energy policies and prediction institutions of energy efficiency are really necessary in each province to improve their energy consumption structures, especially in coordinating the development between the traditional energy and renewable energy.

Due to the complex operational characteristics of multi-level rail transit networks, such as cross-system and multi-level, passenger flow congestion must not only consider the steady state of homogeneous transportation networks but also reveal the deep-seated mechanism of congestion spreading between heterogeneous transportation networks. An analysis theory of travel paths based on Improved Prospect Theory (IPT) is proposed using generalized travel time and congestion degree as dual reference points. By organically integrating passenger travel modes and routes, a two-layer model of passenger travel mode selection based on Nested Logit-Improved Prospect Theory (NL-IPT) is constructed. On this basis, considering key influencing factors such as the stopping scheme, an improved Susceptible-Infected-Recovered (SIR) model of multi-level rail transit passenger flow congestion propagation under bounded rationality conditions is proposed. Taking the multi-level rail transit in Beijing, China, as an example, the propagation process of passenger flow congestion in multi-level rail transit is simulated and analyzed. Through the sensitivity analysis of critical factors such as gain and loss sensitivity coefficient, propagation rate, and recovery rate, the mechanism of the influence of key parameters on passenger flow congestion propagation is revealed. The results show that when the proportion of waiting passengers heading to subsequent stops of the arriving train is greater than or equal to 0.6, there will be slight fluctuations in the initial stage of congestion propagation. When this proportion decreases by 80%, the congestion propagation range decreases by 23.3%. The research provides a reference for the operation plans and management optimization of multi-level rail transit.

In the classification of motor imagery Electroencephalogram (MI-EEG) signals through deep learning models, challenges such as the insufficiency of feature extraction due to the limited receptive field of single-scale convolutions, and overfitting due to small training sets, can hinder the perception of global dependencies in EEG signals. In this paper, we introduce a network called EEG TBTSCTnet, which represents Three-Branch Temporal-Spatial Convolutional Transformer. This approach expands the size of the training set through Data Augmentation, and then combines local and global features for classification. Specifically, Data Augmentation aims to mitigate the overfitting issue, whereas the Three-Branch Temporal-Spatial Convolution module captures a broader range of multi-scale, low-level local information in EEG signals more effectively than conventional CNNs. The Transformer Encoder module is directly connected to extract global correlations within local temporal-spatial features, utilizing the multi-head attention mechanism to effectively enhance the network’s ability to represent relevant EEG signal features. Subsequently, a classifier module based on fully connected layers is used to predict the categories of EEG signals. Finally, extensive experiments were conducted on two public MI-EEG datasets to evaluate the proposed method. The study also allowed for an optimal selection of channels to balance accuracy and cost through weight visualization.

With psoriasis and eczema being the two most common diseases worldwide, achieving automatic diagnosis could be useful for healthcare cyber-physical system. However, creating such an automatic classification system is still challenging since it cannot learn positional and spatial information from unstable training. In this paper, we propose a novel abscissa-ordinate focused network (AOFNet) with active label smoothing for the identification of psoriasis and eczema from images. The AOFNet incorporates the developed abscissa-ordinate focused module that focuses on abscissa-ordinate information and leverages the attention mechanism to enhance the network’s ability to learn positional and spatial details, resulting in improved classification performance. Additionally, the adoption of an active label smoothing approach effectively mitigates the problem of overconfidence and effectively captures the dynamic changes that occur during training, thereby providing an added boost to the overall performance of the network. To evaluate the proposed healthcare cyber-physical system, extensive experiments are conducted on the clinical psoriasis and eczema dataset, and the results demonstrate that the designed system could gain comparable classification performance.

The accurate diagnosis of transmission line fault types is a prerequisite for quickly removing faults and restoring power supply, as well as the key to effectively reducing user economic losses, ensuring stable operation of the power system. The rapid development of artificial intelligence technology has been a promising way for fault diagnosis. However, the existing methods are still limited by small simples and accuracy of generalization. To overcome these problems, a transmission line fault diagnosis method based on an improved multiple SVM (MSVM) model is proposed in this paper. Firstly, the transmission line was selected as the research object, and its fault types and causes were analyzed in detail. Then, typical fault information are selected and corresponding MSVM model is established. Meanwhile, genetic algorithm (GA) is adopted to optimize model parameters to improve the accuracy of generalization. Finally, an improved IEEE-30 node test system and a real-world testing data are adopted to verify the accuracy and feasibility of the proposed method. Through analysis, fault diagnosis accuracy of the proposed method can be improved by up to 11% with better fitness value.

A low-cost inertial navigation system (INS) and global navigation satellite system (GNSS) fusion position estimator is affected by accuracy limitations and multiple noises, leading to significant errors in positioning estimation. This paper proposes a new fusion algorithm, SSA-ESKF, which combines singular spectrum analysis (SSA) and an error-state Kalman filter (ESKF). The low-cost inertial measurement unit (IMU) and GNSS data obtained from the GNSS receivers are separately subjected to SSA noise reduction. The SSA noise-reduced data is then utilized in the ESKF. Consequently, the SSA-ESKF demonstrates superior performance in terms of lower state errors compared to the conventional ESKF. This approach helps minimize the impact caused by neglecting higher-order terms in the Taylor expansion and enhances the linearization of the ESKF, thereby achieving improved positioning accuracy. However, the SSA typically relies on empirically selecting constant singular values, which may result in an incomplete or excessive separation of signal and noise. To address this limitation, we further propose an adaptive spectral singularity analysis (ASSA) that yields better results when integrating the noise-reduced data into the ESKF. To verify the proposed method, the KITTI dataset experiments and real vehicle experiments with low-cost INS/GNSS were designed. The comparison of experimental results between the KF, ESKF and the SSA-ESKF, ASSA-ESKF indicates the superiority of the ASSA-ESKF. In addition, the ablation experiments were conducted to verify the effectiveness of the SSA on IMU data and GPS data independently, and the results showed the effectiveness of the SSA on GPS data.

High-tech enterprises have obvious knowledge-intensive and technical dense characteristics, and human capital plays an important role in the process of corporate knowledge creation. High-tech enterprise manpower capital stock, flow, and investment three perspectives have built the simulation model of high -tech enterprises human capital of the human capital of high-tech enterprises, using system simulation methods, based on the VENSIM PLE simulation environment, revealing the innovative consortium, high-tech enterprise manpower The impact of capital on innovation performance. The research found that the contribution of human capital stock to innovative performance is the greatest, of which technical human capital has the most impact on the output of innovative performance; and the flow of human capital flow and investment are positive to promote innovative performance output, and the level of contribution is basically the same.

In the era of the knowledge economy, innovation and R&D has been paid more and more attention by people. The importance of the R&D team in improving the innovation performance of an enterprise is self-evident. Furthermore, due to the unique human capital characteristics of the R&D team members, fractal research on the human capital of the R&D team is very necessary. On this basis, this paper builds the human capital fractal model of the R&D team, uses the Vensim PLE simulation analysis environment for simulation analysis, and draws the research conclusions: the R&D team’s desire to improve human capital, the R&D team’s desire to integrate knowledge, and the R&D team’s human capital increase.The increase in the amount of R&D team will positively affect the human capital of the R&D team, however, the change of the decrease in the amount of the human capital of the R&D team will negatively affect the human capital of the R&D team.

In this paper, a 94 GHz dual-polarization high-isolation cassegrain antenna feed by a diagonal horn antenna is designed. The antenna consists of a D = 540 mm main reflector, sub-reflector, and dual-port dual-polarization diagonal horn feed antenna. To obtain similar feed horn E-H radiation patterns, gain, high polarization, and port isolation, a dual-port diagonal horn was designed in the feed. The measured antenna gains are 50.85 dBi and the side-lobe levels are −24.5 dB at E-H radiation patterns of two ports at 94 GHz, respectively. The VSWRs of port 1 and port 2 are less than 1.5:1 in the range 93.2–95.3 GHz, which meets the working requirements of frequency modulated continuous wave and pulse cloud radars. And the isolation between the two ports is above −50 dB, whereas the polarization isolation of each port is above 40 dB. The proposed W-band antenna is a suitable candidate for the dual-polarization cloud radars.

The brightness of low-temperature polycrystalline silicon and metal-oxide (LTPO) active-matrix organic light-emitting diode (AMOLED) display under low-refresh-rate driving condition tends to increase during pixel voltage holding time of each frame even though there is no critical leakage current path from the storage capacitor. We have found that the main cause of the brightness increase is electron trapping into the gate insulator of driving thin-film transistor (DTFT) which operates in the saturation mode prone to hot carrier injection. We propose a novel driving method to suppress the brightness variation by applying a high voltage to the source and drain of the DTFT during the non-emitting time which is inserted periodically for dimming the organic light-emitting diode (OLED) display. Our experimental results show that the brightness variation can be reduced from 3.7 % to 1.4 % at the 63rd gray level by the proposed DTFT reset scheme.

Radar pulse deinterleaving is the core part of electronic reconnaissance system, which is responsible for sorting out the pulses coming from different emitters. In the process of signal deinterleaving, parameter clustering is an important method. The features for clustering are mainly derived from pulse parameters emitted by the radiation source transmitter. When the multiple target signals are seriously mixed in a complex environment, the clustering accuracy decreases. To solve this problem, a distributed clustering method based on spatial information is proposed. First, a multi-node distributed system is built, one of the nodes is taken as the central node, and the pulse information transmitted by other nodes through the link is received. All pulses at the central node are jointly clustered according to the three parameters of radio frequency (RF), pulse width (PW) and time of arrival (TOA), forming an associated pulse description word (APDW). Then, the associated pulse is spatially marked based on its spatial information within the range of the target area of interest. Finally, the associated pulse is re-clustered using the marked information and pulse parameters to achieve effective target separation. The simulation results show that the method can overcome the influence of parameter agility and make full use of the spatial information of the target to improve the accuracy of pulse clustering.

The design of railway alignment and station locations involves two intertwined problems, which makes it a complex and time-consuming task. Especially in mountainous regions, the large 3-dimensional (3D) search spaces, complex terrain conditions, coupling constraints and infinite numbers of potential alternatives of this problem pose many challenges. However, most current optimization methods emphasize either alignment optimization or station locations optimization independently. Only a few methods consider coordinated optimization of alignment and stations, but optimize them sequentially. This paper proposes a concurrent optimization method based on a 3-dimensional distance transform algorithm (3D-DT) to solve this problem. It includes the following components: (1) To optimize the location of stations within specified spacing intervals, a novel perceptual search strategy is proposed and incorporated into the basic 3D-DT optimization process. (2) A combined-alignment-station 3D search neighboring mask is developed and employed to search for both the alignment and stations. In order to implement the perceptual process, two additional kinds of 3D reverse perceptual neighboring masks are also developed and employed in the algorithm. (3) Multiple coupling constraints between alignment and stations are also formulated and addressed during the search process. In this study, the effectiveness of the method is verified through a real-world case study in a complex mountainous region. The optimization results show that the proposed method can find high-quality alternatives satisfying multiple coupling constraints.

A novel H -plane waveguide in-phase power divider (also known as combiner) with high isolation working across full WR-3 band (220-330 GHz) is realized in this paper. This distinctive architecture is designed based on the sum-difference principle of Magic-T, whose E -plane branch (difference port) is matched with an E -plane T-junction. Both effects of processing and assembling uncertainties on this power divider/combiner performance are discussed in detail to display error sensitivity. The physical prototype is split into three blocks and processed by the state-of-the-art computer numerical control (CNC) milling technology. The absorbing materials are housed in the loaded T-junction arms to terminate return wave as well. Compared with the most reported terahertz waveguide power couplers, the meaningful measured results of this advanced in-phase power divider are delivered. The input and output matching as well as high isolation have been obtained over full WR-3 band, when an 11 dB reference level is considered. A total insertion loss of about 1.2 dB is achieved including three straight waveguides having a total length of 30 mm, which is also in good agreement with simulations. This module, which is fully compatible with the standard CNC-milling technique, can be directly integrated to terahertz detection systems.

Filters play an invaluable role in most detection and communication systems, blocking unwanted, harmonic or mirrored waves. In this paper, two $3^{\mathrm {rd}}$ -order quasi-elliptical waveguide bandpass filters working at 400 GHz are developed based on TE102-mode oversized resonators. Transmission zeros are generated through the physical cross-couplings and the mixed modal bypass couplings. $H$ -plane off-axis coupling structures without any fragile irises are adopted to strengthen robustness and simplify manufacturing at such high terahertz frequencies. Effects of processing uncertainties on filters performance are discussed in detail to display error sensitivity. Two computer numerical control (CNC) machined filters exhibit a minimum insertion loss of about 1.5 dB in a 3 dB fractional bandwidth (FBW) with 12% centered at 390 GHz @ Filter-I, and 9% FBW with a center frequency of 394 GHz @ Filter-II, which are all mainly agreement with simulations. The performance is highlighted with the reported terahertz filters, which indicates that both architectural designs are still suited for the delicate CNC-milling restriction in such high WR-2.2 band.

Audio steganography aims to exploit the human auditory redundancy to embed the secret message into cover audio, without raising suspicion when hearing it. However, recent studies have shown that the existing audio steganography can be easily exposed with the deep learning based steganalyzers by extracting high-dimensional features of stego audio for classification. The existing GAN-based steganography approaches mainly studied in images cover, less work is conducted on audio cover. In addition, though a few GAN-based audio steganography methods have been proposed, they still have room for improvements in perceptual quality and undetectability. In this work, we propose an audio steganography framework that could automatically learn to generate superior steganographic cover audio for message embedding. Specifically, the training framework of the proposed framework consists of three components, namely, generator, discriminator and trained deep learning based steganalyzer. Then the traditional message embedding algorithm LSBM, is employed to embed the secret message into the steganographic cover audio to obtain stego audio, which is delivered to the trained steganalyzer for misclassifying as cover audio. Once the adversarial training is completed among these three parties, one can obtain a well-trained generator, which could generate steganographic cover audio for subsequent message embedding. In the practice of our proposed method, the stego audio is produced by embedding the secret message into the steganographic cover audio using a traditional steganography method. Experimental results demonstrate that our proposed audio steganography can yield steganographic cover audio that preserves a quite high perception quality for message embedding. We have compared the detection accuracies with the existing audio steganography schemes as presented in our experiment, the proposed method exhibits lower detection accuracies against the state-of-the-art deep learning based steganalyzers, under various embedding rates. Codes are publicly available at https://github.com/Chenlang2018/Audio-Steganography-using-GAN.

Modern vehicles are equipped with more than 100 Electrical Control Units (ECUs) with over 2500 signals to transmit internally. The application of advanced electronics and communication techniques helps a vehicle transform from an information island into a powerful distribution center. However, a large number of ECUs have introduced a wider range of security threats for vehicles. The attackers can compromise a vehicle remotely through a vulnerable ECU. How to evaluate the cyber security of in-vehicle ECUs has become an important issue. Current Threat Analysis and Risk Assessment (TARA) only carries out theoretical analysis on the potential threats and risks faced by the vehicle in the conceptual design phase of the lifecycle, but lacks the details of actual security evaluation. In this paper, we proposed a Cyber Security Evaluation Framework (CSEF) to independently evaluate the security of the in-vehicle ECUs, which is composed of the asset identification, the threat analysis, the risk assessment, and the security test. The proposed CSEF is applied to a pre-installed On-Bord Unit (OBU) to provide a use case. The use case show that the proposed CSEF is able to figure out assets, threats, risks behind threats, and vulnerabilities of OBU, playing an important role in guiding others to conduct security evaluation. Moreover, CSEF can be extended to evaluate the cyber security of other critical ECUs, such as the Telematic Box, the infotainment units, and the gateway.

Dry bulk shipping plays an essential role in international bulk commodity transportation. The dry bulk shipping company that operates a fleet has to design an efficient and safe route for each vessel to complete the transportation task and reduce the navigation risk on the course. Considering the navigation risk is uncertain, we use the robust optimization approach to optimize the routes for a dry bulk shipping fleet. We aim to determine the shipping route that can minimize the total transportation cost and ensure that the navigation risk is no more than a certain threshold. To address this problem, this paper first develops a deterministic formulation for the dry bulk fleet route planning. Then its robust counterpart is formulated to incorporate the uncertain navigation risk, which can ensure the worst-case navigation risk shall not exceed the given threshold. At last, we carry out a series of numerical experiments using actual data to evaluate the performance of our approach. The numerical results reveal that our approach can efficiently solve the route planning problem and make the trade-off between navigation risk and total transportation cost.

Compressed sensing has shown great potential and power in image representation, especially in image reconstruction by sparse representation. Due to complementary information and unavoidable noise existing in synthetic aperture radar (SAR) and other source images, joint sparse representation (JSR) is developed to separate redundancy and complementary information with different properties in source images and obtain a fused image, where image de-noising is done simultaneously owing to that noise is not sparse and cannot be represented by sparse representation. As a result, one noisy remote sensing image fusion method based on JSR is presented in this paper. After obtaining redundant and complementary sub-images by JSR, an improved fusion rule based on pulse coupled neural network (PCNN) is employed to fuse complementary sparse coefficients together. At the same time, because the types of noise in SAR and other source images are different, they can be treated as the complementary information in source images and suppressed at this step. Finally, a fused image can be reconstructed by adding the redundant and fused complementary sub-images. Quantitative and qualitative experimental results show that the proposed method outperforms most of other fusion methods and it is more robust to noise, having better visual effects and values of objective evaluation metrics.

Acquired Immune Deficiency Syndrome (AIDS) is still one of the most life-threatening diseases in the world. Moreover, new infections are still potentially increasing. This difficult problem must be solved. Early warning is the most effective way to solve this problem. Here, we aim to determine the best performing model to track the epidemic of AIDS, which will provide a methodological basis for testing the time characteristics of the disease. From January 2004 to January 2018, we built four computing methods based on AIDS dataset: BPNN model, RNN model, LSTM model and MHPSO-GRU model. Compare the final estimated performance to determine the preferred method. Result. Considering the root mean square error (RMSE), mean absolute error (MAE), mean error rate (MER) and mean absolute percentage error (MAPE) in the simulation and prediction subsets, the MHPSO-GRU model is determined as the best performance technology. Estimates for the period from May 2018 to December 2020 suggest that the event appears to continue to increase and remain high.

To improve the detection of dim and small targets in infrared (IR) images containing high-intensity cloud clutter, a novel adaptive background suppression method is proposed. By using three-dimensional cooperative filtering and differential calculation, the different and complex background clutter is suppressed. To obtain the optimal parameters for the background suppression algorithm, an adaptive parameter optimization method is proposed. The adaptive parameter optimization problem is transformed into a multiobjective optimization problem in which the signal-to-clutter ratio gain and background suppression factor, which effectively reflect the background suppression performance, are chosen as the optimization objectives, and the parameters of the proposed background suppression algorithm are considered as the variables. To effectively solve the established multiobjective optimization problem, a particle swarm parameter optimization-based method is utilized. Experimental results indicate that the proposed adaptive background suppression method using these optimal parameters has good performance for IR images in real complex scenes, as well as performance superior to that of other baseline methods.

Although neural network-based speech recognition models have enjoyed significant success in many acoustic systems, they are susceptible to be attacked by the adversarial examples. In this work, we make first step towards using generative adversarial network (GAN) for constructing the targeted speech adversarial examples. Specifically, we integrate the target speech recognition network with GAN framework, which can then be formulated as a three-party game. The generator in GAN aims at generating perturbation that could make the target network misclassified to a specific target, while simultaneously fooling the discriminator treating the adversarial example as a beguine one. The discriminator is to distinguish the crafted examples from the geniue samples. The classification error of the target network is back-propagated via gradient flow to the generator for updating. The target network is responsible for back-propagating the classification error via gradients to the generator for updating, but the target network itself is freezed. With the carefully designed network architecture, loss function and training strategy, we successfully train a generator that could generate the adversarial perturbation for a given speech clip and a target label. Experiential results show that the generated adversarial examples could effectively fool the state-of-the-art speech classification networks, while attaining an acceptable auditory perception quality. In addition, our proposed method runs much faster than the prevalent optimization-based schemes. To facilitate reproducible research, codes, models and data are publicly available at https://github.com/winterwindwang/SpeechAdvGan.

Recently, Siamese trackers have shown excellent performance in both accuracy and speed. However, traditional trackers have poor robustness against similar objects due to the use of single deep features and the limitation of cosine windows. In this paper, a novel Siamese network combining information fusion with rectangular window filtering named SiamFF is introduced. First, a multilevel fusion network is proposed. At feature-level, the shallow and deep features of the network are fused through a layer-hopping connection to obtain complementary feature maps. Then, the score maps generated by the complementary feature maps are further fused at the score-level to improve the robustness. In addition, based on the continuity and stationarity of objects movement in reality, a score map filtering strategy is proposed. The relative displacement of the target can be predicted by obtaining the interframe information, and the moving direction is applied to filter the score map to further eliminate the analog interference. Experimental results on OTB2015 and VOT2016 benchmarks indicate that SiamFF performs favorably against many state-of-the-art trackers in terms of accuracy while maintaining real-time tracking speed.

This paper reports the investigation on non-fragile state feedback control for stochastic Markovian jump systems with uncertain parameters and mode-dependent time-varying delays. The resulted closed-loop system is stochastic stabilization by virtue of an improved L-K functional. By free-weight-matrix technique, the non-fragile state feedback controller is designed and novel conditions for robust stochastic stabilization are acquired in the form of linear matrix inequalities. Two examples including a take-off and landing guidance system are employed to show the effectiveness and validity of the proposed approach.

As the distributed generation (DG) in a power supply and the user load demand constantly change in an actual distribution network, multiobjective optimal network reconfiguration considering variations in load and DG has become a major concern, which is important and required to make system operations safe and economical. The aim is to minimize the sum of the active power loss, the sum of the load balancing index and the sum of the maximum node voltage deviation index simultaneously during the reconfiguration period. Here, this article proposes a new Chaos Disturbed Beetle Antennae Search (CDBAS) algorithm to reduce the computational time and solve the multiobjective optimal problem of network reconfiguration. To adopt the Chaos Disturbed Beetle Antennae Search algorithm for solving this multiobjective problem, grey target decision-making technology is used to rank the beetles. Additionally, to the enhance the system static voltage stability and voltage quality, a grey target decision-making model is established to achieve a layer relationship between each index and the switching operation index. The plausibility and effectiveness of the presented methodology is verified on the modified IEEE 33, 69 and 118-Bus Test Radial Distribution Network. Finally, compared with other research methods in the literature, the CDBAS algorithm outperforms other algorithms and produces a quality decision solution.

Considering the problems of motion blur, partial occlusion and fast motion in target tracking, a target tracking method based on adaptive structured sparse representation with attention is proposed. Under the framework of particle filtering, the performance of high-quality templates is enhanced through an attention mechanism. Structure sparseness is used to build candidate target sets and sparse models between candidate samples and local patches of target templates. Combined with the sparse residual method, reconstruction error is reduced. After optimally solving the model, the particle with the highest similarity is selected as the prediction target. The most appropriate scale is selected according to the multiscale factor method. Experiments show that the proposed algorithm has a strong performance when dealing with motion blur, fast motion, partial occlusion.

In this paper, we propose a transfer learning-based dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) method for classifying fibroadenoma and invasive ductal carcinoma (IDC) in breast tumors. A total of 207 breast tumors from patients were collected and identified by pathologic diagnosis within 15 days after enhanced DCE-MRI examination; 119 patients (average age 50.52±10.33 years) had fibroadenomas, and 88 patients (average age 42.20±10.10 years) had IDCs. Two lesion-level models were built based on the InceptionV3 and VGG19 models, which were pretrained with the ImageNet dataset. The effects of different depths of transfer learning were examined. The network's performance was evaluated through five-fold cross validation. In the lesion-level models, the model based on Inception V3 obtained better results (area under the receiver operating characteristic curve (AUC) = 0.89) when the weights were frozen before layer-276. The other model based on VGG19 obtained better results (AUC = 0.87) when the weights were frozen before layer-13. Compared with the image-level models, both lesion-level models displayed better discrimination (accuracy increased by 13% and 14%) based on the VGG19 and Inception V3 models, respectively. Our research confirms that transfer learning can be utilized to classify fibroadenomas and IDCs in DCE-MRI images. Different depths of transfer learning result in different performances, and our proposed lesion-level model notably improves the classification accuracy.

Aiming at the problem that the petroleum industry is a high-occurrence industry with many occupational hazards, in order to comprehensively evaluate the hazards in coal mines, the number of supervisors for petroleum occupational hazards is small, the number of objects to be supervised, the frequency of supervision, and the scope of supervision are large, combined with cloud computing Based on the characteristics of cloud computing, an occupational hazard supervision system for petroleum companies based on cloud computing is proposed. Based on relevant theories, 15 occupational hazard identification indicators such as lifestyle, average working age, cultural level, professional skills, and health level were selected from massive data to establish an occupational hazard evaluation index system. In this paper, the analytic hierarchy process is used to determine the weight of the occupational hazard identification index, and then a BPNN neural network prediction model is established based on the determined weight index system. The qualitative prediction of the input and output of the index is achieved through model solving and model training. The experimental results show that the training error accuracy of BPNN neural network is less than 5%, and the model accuracy meets the requirements. Compared with the error of AHP, the average evaluation error of BPNN is 0.637% lower than that of AHP. The comprehensive evaluation system based on BPNN accords with the actual situation of the enterprise is greater than 97%, so it has certain guiding significance in the evaluation of petroleum occupational diseases.

With the increased distributed generation (DG) and electric vehicle (EV) load penetration in distribution networks, it is more difficult to ensure the safe and economic operation of the distribution networks because of the great volatility and randomness of DG and EV loads. In this article, the uncertainties of wind power, photovoltaics, conventional loads, and EV loads are considered. Photovoltaics and conventional loads are related to solar radiation, and they are subtracted to form the net load. Then, the Wasserstein distance is used to divide the scene, and K-means clustering is used to reduce the scene, so the reconstruction analysis is carried out in the limited typical scene. In addition, to minimize network loss, load balance and maximum voltage deviation, a multiobjective game reconstruction model of the distribution network is established under the condition of satisfying network constraints. Moreover, the social beetle swarm optimization algorithm considering two social behaviors is adopted to solve the complex problem. Finally, the simulation results are verified on the standard IEEE-33 system and IEEE-118 system. The results show that the proposed strategy and algorithm can effectively reduce the network loss, improve the node voltage level and ensure that the load is not overloaded.

Among the accidents in China's underground noncoal mines including ferrous metal mines, nonferrous metal mines and nonmetallic mines, roof fall accidents are always extremely bad. The application of risk assessment to avoid roof fall accidents in coal mines is widespread. However, the traditional method has defects in its accuracy, cost, and simplicity. It cannot fundamentally ensure that workers survive a roof fall accident. Therefore, there is no great referential significance for noncoal mines. In this paper, supposing that roof fall accidents in underground noncoal mines are inevitable, we assess where we should start to take measures to prevent worker deaths. On the basis of a detailed investigation, units including those working far from the roof, noticing the signs, etc. were identified, and then a concise event tree model was built. Under this framework, the event tree model has 2 consequences, including being alive and death, and 13 scenarios related to accident occurrence. In this research, the probability of unit events was evaluated by using triangular fuzzy numbers. In addition, we found 3 scenarios with death, and then we calculated their respective probabilities. Finally, regarding the death scenario with the highest risk value, the paper recommends that priority measures should be taken to prevent deaths in roof fall accidents. The method in this paper provides a theoretical basis for pre-control measures and it is efficient, inexpensive, and simplified. Even if a roof fall accident is inevitable, we know where to start to take pre-control measures to prevent worker deaths.

Aiming to improve the operation economy and power quality of distribution systems subject to the fluctuating and stochastic power outputs of distributed generation (DG) units and electric vehicles (EVs), a multiobjective optimization model for network reconfiguration and its corresponding solution method are proposed. First, a dynamic reconfiguration model is constructed based on the power loss rate (PLR) as well as the active power loss, the load balancing index and the maximum node voltage deviation, which serve as the optimization indexes. Second, the Lévy flight and chaos disturbed beetle antennae search (LDBAS) algorithm is presented based on the grey target decision-making technique, which can not only improve the computational efficiency but also find the most satisfactory solution for the proposed dynamic reconfiguration model. Considering the uncertainties of loads and DG outputs, the influence on the load curve of EV connecting to the distribution network at different penetration rates and under different charging/discharging modes is analysed. Additionally, the modified IEEE 33-bus and 118-bus test radial distribution networks are simulated to verify the effectiveness and superiority of the presented reconfiguration model and the improved LDBAS method, and the results illustrate that the proposed reconfiguration method can improve the operation economy and power quality of the distribution system and encourage the penetration of EVs.

In modern chemical process control, the application of data-driven soft sensor has become increasingly extensive. Feature extraction is an important step in soft sensor. A novel feature extraction and integration method based on stacked autoencoders (SAE) and mutual information (MI)-weighted principle component analysis (PCA) was proposed to solve the loss of information on shallow depth features and original variables in neural network models. First, an SAE model was trained to extract the features of the original variables with varying depths. Second, through an MI indicator, the original variables and features with strong dependency on the outputs were selected. Then, MI was used to assign varied weights to the features and original variables, and the PCA method was used to remove any possible redundancy between the original variables and features of varying depths to obtain the principle components. Finally, the principle components were used to construct a regressor, such as a neural network. The model was first tested using the Boston housing dataset as a benchmark and then applied to the soft sensor of a constant top oil dry point. The proposed model achieved optimal results in terms of the root mean squared error and $r$ indicators in the experiments and was thus proved feasible and useful.

Fault-tolerant systems are indispensable for complex electronic systems, but there are some drawbacks to the traditional fault-tolerant schemes. The evolution of hardware for fault-tolerant systems can achieve good results. However, because of the highly time-consuming evolution process, the evolution approach may not meet the real-time constraint for fault recovery. In this paper, we propose an improved genetic algorithm for real-time circuit fault recovery. We establish a real-time fault-tolerant system on a ZYNQ chip. In this system, a fault analysis tree hierarchically monitors a circuit fault. When the fault occurs, the normal operation of the faulty circuit is temporarily maintained by a fault compensation mechanism. At the same time, we use an evolutionary mechanism combined with a fault recovery library and improved genetic algorithm to accelerate the evolution of the repair circuit and obtain a repaired circuit while the fault compensation mechanism is running. Ultimately, the improved algorithm significantly improves the fault-tolerant recovery rate and ensures the operations in real time. Thus, the improved genetic algorithm for real-time circuit fault recovery can meet the system's real-time constraints and improve the system's stability.

A robust controller taking advantage of feedback linearization control (FLC) and sliding mode control (SMC) is proposed with an objective of mitigating the sub-synchronous control interaction (SSCI) in series-compensated doubly fed induction generator (DFIG)-based wind farms (WFs). FLC method is employed to obtain the reduced-order model of the studied system and, hence, smaller computation burden to controller design. Moreover, unlike linear control methods, FLC enables the controlled system to be independent of the pre-specified operations, which is suitable for highly nonlinear DFIG-based WFs. Considering that the FLC can be sensitive to parameter uncertainties, SMC is combined with the FLC to improve the robustness of the system. To evaluate the performance of the proposed feedback-linearized sliding mode controller (FLSMC) as compared to the conventional sub-synchronous resonance damping controller, the electromagnetic transient simulation and small-signal stability analysis are carried out. The designed FLSMC is observed to demonstrate the effectiveness in SSCI mitigation and robustness against the parameter perturbation at varied operating conditions. The experimental tests are performed to validate the veracity of the simulation results.

A home-built scattering-type scanning near-field millimeter-wave microscope based on a 110-GHz continuous-wave solid-state source is demonstrated with a spatial resolution of 1 μm, approximately 1/3000 of the incident wavelength, and a signal-to-noise ratio of 23.8 dB. The relationship between the length of the tip (antenna) and the wavelength for resonant enhancement, and the near-field distribution around the tip apex at different tip-sample separations were explored using finite-difference time-domain electromagnetic simulations to facilitate the design of the microscope. The dependence of the spatial resolution on the tip-sample separation and the harmonic order of the modulation frequency at which the near-field signal was extracted has been investigated experimentally and discussed in terms of the signal-to-noise ratio and the standard deviation of the demodulated signal.

To get a better fused performance in the multi-focus image fusion based on a transform domain, a new multi-focus image algorithm combined with the adaptive dual-channel spiking cortical model (SCM) in non-subsampled shearlet (NSST) domain and the difference images is proposed in this paper. First, a basic fused image is constructed in the NSST domain by registering the source image and adaptive dual channel SCM (dual-channel SCM). Next, the focus areas of the sources input images based on the difference images between the basic fused image and the sources images are detected. In the end, the final fused image generated in this paper is realized by combining the focal regions. Because of the global coupling of the dual-SCM, the synchronization characteristics of the pulse, and the multi-resolution and direction of the NSST, the proposed algorithm can preserve the information of the source's image well and present a clear image more in line with the human visual effects. In summary, the image fusion algorithm that we have designed is superior to the most advanced algorithms.

Recently, the permanent magnet linear synchronous motor (PMLSM) servo control technology has become a research hotspot. However, the control performance under the traditional PID control is not ideal, which usually has the problems of slow response speed, large overshoot and poor anti-interference performance. In addition, adding nonlinear links with certain characteristics purposefully can greatly improve the performance of the system and achieve the expected effect that the traditional linear controllers cannot achieve. Based on this, this paper introduces the nonlinear PID control strategy into the field of servo control, and proposes a nonlinear control strategy of the three-loop of the servo system to improve the above problems. Firstly, the proposed structure is explained and the rule of nonlinear parameters selection is discussed. Secondly, the conditions for the structural stability of the proposed nonlinear control strategy are analyzed. Finally, experiments on the servo system platform are carried out and the results verify the fast dynamic response, precise positioning and good anti-disturbance performance.

In this paper, a finite-time control strategy composed of disturbance observer and sliding mode controller is investigated for five-degree of freedom (5-DOF) upper-limb exoskeleton subject to system uncertainties and external disturbances. First, a new finite-time disturbance observer (FTDO) is constructed to handle unknown disturbances and modeling uncertainties. Second, a nonsingular fast terminal sliding mode (TSM) approach is presented to follow the desired joint angles. By combining the novel FTDO and the nonsingular fast TSM technology, the exoskeleton angle tracking errors can be forced to zero within a finite time. Furthermore, the stability of the exoskeleton system is illustrated via Lyapunov. Finally, the tracking performance of the proposed control scheme is verified by simulations.

In order to obtain a panoramic image which is clearer, and has more layers and texture features, we propose an innovative multi-focus image fusion algorithm by combining with non-subsampled shearlet transform (NSST) and residual network (ResNet). First, NSST decomposes a pair of input images to produce subband coefficients of different frequencies for subsequent feature processing. Then, ResNet is applied to fuse the low frequency subband coefficients, and improved gradient sum of Laplace energy (IGSML) perform high frequency feature information processing. Finally, the inverse NSST is performed on the fused coefficients of different frequencies to obtain the final fused image. In our method, we fully consider the low frequency global features and high frequency detail information in image by using NSST. For low-frequency coefficients fusion, we can also obtain the spatial information features of low-frequency coefficient images by using ResNet, which has a deep network structure. IGSML can use different directional gradients to process high-frequency subband coefficients of different levels and directions, which is more conducive to the fusion of the coefficients. The experiment results show that the proposed method has been improved in the structural features and edge texture in the fusion images.

On account of the indeterminacy and subjectivity of decision makers (DMs) in complexity decision-making environments, the evaluation information over alternatives presented by DMs is usually fuzzy and ambiguous. As the generalization of intuitionistic fuzzy sets (IFSs) and Pythagorean fuzzy sets (PFSs), the q-rung orthopair fuzzy sets (q-ROFSs) are more useful to express more fuzzy and ambiguous information. Meanwhile, to consider human's hesitance, the dual hesitant q-rung orthopair fuzzy sets (DHq-ROFSs) are presented which can be more valid of handling real MADM problems. To fuse the information in DHq-ROFSs more effectively, in this article, some Muirhead mean (MM) operators based on DHq-ROFSs environment, which consider any number of being fused arguments, are defined and studied. Evidently, the new proposed operators can obtain more exact results than other existing methods. In addition, some precious properties of these MM operators are discussed and all the special cases of them are investigated which indicates MM operator is more powerful than others. Afterward, the defined aggregation operators are used to solve the MADM with dual hesitant q-rung orthopair fuzzy numbers (DHq-ROFNs) and the MADM decision-making model is developed. In accordance of the defined operators and built model, two operators are applied to deal with the MADM problems for supplier selection with the DHPFNs information and the availability and superiority of the proposed operators are analyzed by comparing with some existing approaches. The method presented in this paper can effectually solve the MADM problems in which the decision-making information is expressed by the DHq-ROFNs and the attributes are interactive.

As we all know, the Bonferroni mean (BM) operator has the advantage of considering interrelationships between parameters. In this paper, we combine the generalized weighted BM (GWBM) operator and generalized weighted Bonferroni geometric mean (GWGBM) operator with interval neutrosophic numbers (INNs) to develop the generalized interval neutrosophic number weight BM (GINNWBM) operator and generalized interval neutrosophic numbers weighted GBM (GINNWGBM) operator which consider the relationship among three aggregated arguments, then the MADM methods are developed with these operators. Finally, we use an example for evaluating the technological innovation capability for the high-tech enterprises to illustrate the proposed methods.

Existing video steganography puts much emphasis on the design of algorithms such as mapping rules or distortion functions, thereby ignoring the selection of cover to embed secret information. However, this is just one of the major differences between image steganography and video steganography. In addition, since HEVC is the latest standard in the video codec field, it is of important academic significance and applied value to study HEVC-based steganography. This paper proposes a novel video steganography in HEVC, based on intra-prediction mode (IPM). Firstly, this paper analyzes the probability distribution of 4 × 4 IPMs. Then a cover selection rule combined with the Coding Unit (CU) and Prediction Unit (PU) coding information is proposed, which can improve the security performance of a stego video stream. In addition, matrix coding is used as a coding example to implement the steganography on HEVC video streams. Experimental results show that the proposed algorithm can not only maintain the video quality and the security performance but is also easy to implement. Furthermore, the proposed cover selection rule can also be integrated into other HEVC IPM based steganography.

In this study, a three-degree-of-freedom hydraulic system device configured as a master-and-slave manipulator system is considered. A methodology to estimate reaction force in aforementioned system is proposed without utilizing any force sensor/signal. A terminal sliding control with a sliding perturbation observer (TSMCSPO) is designed for coupled master-slave system. The fundamental benefit of the proposed scheme is that the sliding perturbation observer estimates the reaction force related to the end effector and second link without utilizing a force sensor. Robust and accurate position tracking is achieved with terminal sliding-mode control in addition to control of the slave manipulator. The position and force tracking for master-slave is estimated based upon bilateral control strategy. The difference in the reaction and operating force is included in the impedance model. The impedance model is implemented in the master device to provide the operator with a sense of the reaction force when it connects with the environment. The experimental results confirm the robust position and force tracking of the slave manipulator.

The steel industry is one of the pillar industries in China. The physical and mental health of steel workers is related to the development of China's steel industry. Steel workers have long been working in shifts, high temperatures, noise, highly stressed, and first-line environments. These occupational related factors have an impact on the health of steel workers. At present, the existing hypertension risk scoring models do not include occupational related factors, so they are not applicable to the risk score of hypertension in steel workers. It is necessary to establish a risk scoring model for hypertension in steel workers. In this study, the learning vector quantization (LVQ) neural network algorithm and the FisherSVM coupling algorithm are applied to estimate the hypertension risk of steel workers, and the microscopic laws of the "tailing" phenomenon of the two algorithms are analyzed by means of graphics analysis, which can describe the influence trend of sample size change in different intervals on the classification effect. The results show that the classification accuracy of the algorithm depends on the size of the sample space. When the sample size n ≤ 30 * (k + 1), the Fisher-SVM coupling intelligent algorithm is more applicable. Because its average accuracy rate is 90.00%, the average accuracy of the LVQ algorithm is only 63.34%. When the sample size is n > 30 * (k + 1), the LVQ algorithm is more applicable. Because its average accuracy rate is 93.33%, the average accuracy of the Fisher-SVM coupling intelligent algorithm is only 76.67%. The sample size of this paper is 4422, and the prediction of LVQ neural network model is more accurate. Therefore, based on the relative importance of each risk factor obtained by this model and to establish a steel worker hypertension risk rating scale, the score greater than 18 is considered as the high risk, 12-18 is considered as the medium risk, and less than 12 is considered as the low risk. Through the example's verification, the accuracy rate of the scale is 90.50% and the effect is very good. It shows that the established scoring system can effectively assess the risk of hypertension in steel workers and provide an effective basis for primary prevention of hypertension in steel workers.

Frequency division duplex (FDD) systems dominate current cellular networks due to its advantages of low latency and strong anti-interference ability. However, the computation and the feedback overheads for predicting the downlink channel state information (DL-CSI) are the major bottlenecks to further improve the cellular FDD systems performance. To deal with these problems, in this paper, a convolutional long short-term memory network (ConvLSTM-net)-based deep learning method is proposed for predicting the DL-CSI from the uplink channel state information (UL-CSI) directly. In detail, our proposed ConvLSTM-net consists of two modules: one is the feature extraction module that learns spatial and temporal correlations between the DL-CSI and the UL-CSI, and the other one is the prediction module that maps the extracted features to the reconstructions of the DL-CSI. To evaluate the outperformance of the ConvLSTM-net, a long short-term memory network (LSTM-net) and a convolutional neural networks (CNN)-based schemes are simulated for comparisons. The simulation experiments consist of two parts. One part is that the hyper parameters of the proposed ConvLSTM-net are analyzed to explore their effects on the prediction performance. Another part is that experiments are conducted in the time domain and frequency domain, respectively, for selecting a more proper domain to predict the DL-CSI accurately. From the experiment results above, it can be verified that the proposed ConvLSTM-net with proper hyper parameters outperforms the compared schemes at predicting DL-CSI according to UL-CSI in the cellular FDD systems, especially in the time domain.

Current artificial intelligence models for evaluating goaf stability in underground metal mines need a large amount of sample data for training, and their accuracy declines with small sample size. With the aim of solving this problem, this paper proposes an improved TrAdaBoost algorithm based on transfer learning theory. The scope of the TrAdaBoost algorithm is extended from the two-level classification to multi-level classification problems, which makes it suitable for evaluating goaf stability. The isolated forest method is used to filter the bad points of the auxiliary training set, thereby eliminating the interference of abnormal data. The dynamic factor concept is introduced to solve the problem that the weight of source domain data decreases too quickly and irreversibly, and this enhances the generalization performance of the algorithm for different goaf samples. To test the accuracy of the proposed model in predicting goaf stability, an evaluation model is constructed and the performance compared with other algorithms in current use. The prediction accuracy and generalization ability of the model are evaluated by mean square error and F1 measurements, which prove that the performance of the model is excellent. The most obvious finding to emerge from this paper is that, with suitable improvements, goaf evaluation models can still maintain a high level of accuracy with small sample size.

The Dombi t-norm and Dombi t-conorm operations, which have the advantages of describing decision making information flexibly by using the general parameter, can be valid to deal with actual multiple attribute decision making (MADM) problems. Motivated by the basic Bonferroni mean (BM) operator, some novel Dombi Bonferroni mean (DBM) operators and Dombi geometric Bonferroni mean (DGBM) operators are developed to aggregate 2-tuple linguistic neutrosophic information in our manuscript. Obviously, some precious properties and numerical examples are given to testify the effective and scientific of these operators. Finally, the green supplier selection in green supply chain management is taken as an example to verify the proposed methods.

To fuse the information in dual-hesitant Pythagorean fuzzy sets (DHPFSs) more effectively, in this paper, some dual hesitant Pythagorean fuzzy Hamy mean (DHPFHM) operators, which can consider the relationships between being fused arguments, are defined and studied. Afterward, the defined aggregation operators are used to multiple attribute decision-making (MADM) with dual-hesitant Pythagorean fuzzy elements (DHPFEs), and the MADM decision-making model is developed. In accordance with the defined operators and the built model, the dual-hesitant Pythagorean fuzzy weighted Hamy mean (DHPFWHM) operator and the dual-hesitant Pythagorean fuzzy weighted dual Hamy mean (DHPFWDHM) operator are applied to deal with green supplier selection in supply chain management, and the availability and superiority of the proposed operators are analyzed by comparing with some existing approaches. The method presented in this paper can effectually solve the MADM problems, which the decision-making information is expressed by DHPFEs and the attributes are interactive.

The linguistic q-rung orthopair fuzzy sets (Lq-ROFSs) can fully represent linguistic evaluation information by adjusting the parameter $q$ to describe the range of uncertain information. In this paper, we first propose a cosine similarity measure between Lq-ROFSs based on the concept of the cosine similarity measure between intuitionistic fuzzy sets. Considering the cosine similarity measure defined by this method does not satisfy the axiom of similarity measure, we propose the improved similarity measures of Lq-ROFSs according to the corresponding cosine similarity measure and Minkowski distance measure. The improved similarity measures of Lq-ROFSs can deal with the decision information not only from the point of view of algebra but also from the point of view of geometry, which are also proved satisfy the axiom of similarity measure. Furthermore, the similarity measures of Lq-ROFSs based on different linguistic scale functions are defined for considering the semantics of linguistic terms. In addition, we apply the proposed similarity measures of Lq-ROFSs to multi-criteria group decision making problems where the decision makers’ weights are determined by the projection of individual decision on the ideal decision results. Finally, a numerical example is applied to illustrate the feasibility of the proposed method and its effectiveness is verified by comparison with some existing methods. The sensitivity analysis and stability analysis of the proposed method are also given.

In this paper, we defined ten similarity measures between spherical fuzzy sets (SFSs) based on the cosine function by considering the degree of positive-membership, degree of neutral-membership, degree of non-membership and degree of refusual-membership in SFSs. Then, we applied these similarity measures and weighted similarity measures between SFSs to pattern recognition and medical diagnosis. Finally, two illustrative examples are designed to show the efficiency of the similarity measures for pattern recognition and medical diagnosis.