Search

Recently, steganalysis of Voice over Internet Protocol (VoIP) compressed speech has gained attention. In real voice communication, Joint Parallel Steganography (JPS) often occurs, where multiple steganography algorithms coexist. The multifaceted nature of JPS, incorporating various steganographic algorithms, poses significant challenges in steganalysis. We believe that detecting JPS accurately requires multi-stage feature extraction, as a single-stage approach fails to yield satisfactory results. In this paper, we propose an efficient steganalysis model based on Inter-codeword Sensitivity Caption, termed InSeC. It consists of two neural modules: the steganography-sensitive codeword-pair caption module, which analyzes changes in codeword pairs before and after modification from multiple perspectives and aggregates these features, and the fine-grained correlation re-perception module, which re-evaluates features within a local range. Our approach improved detection precision by 25.27%, 11.57%, 9.07%, and 9.28% compared to four recent comparison methods on the JPS detection task. The source code for this work is publicly available on https://github.com/zhousandeqingshu/ZhCode.

The time-dependent green vehicle routing problem with time windows is a further deepening of the research on vehicle routing problems with time windows. Its simultaneous consideration of vehicle transportation time, carbon emissions, and customer satisfaction under time-dependent variables makes it more challenging to solve than traditional vehicle routing problems. This work proposes a multi-objective optimization algorithm that combines the learnable crossover strategy and the adaptive search strategy based on reinforcement learning to overcome the local optima, poor convergence, and reduced variety of solutions that plague the multi-objective optimization algorithms when solving this problem. The proposed approach solves the problem in two stages: In the first stage, a hybrid initialization strategy is used to generate initial solutions with high quality and diversity, and crossover strategies are used to further explore the solution space and improve convergence by learning the characteristics of pareto solutions. In the second stage, the adaptive search is designed and used for learning and searching in the later stage of the algorithm. The experimental results show better solution quality obtained by the proposed approach, and the effectiveness and superiority of the proposed approach over existing methods in terms of solution convergence and diversity are demonstrated through experimental comparisons.

The paper proposes a path planning method for meteorological detection of the unmanned aerial vehicle (UAV) in complex obstacle environments based on an enhanced multi-strategy whale optimization algorithm. In the design of the modeling, a weight coefficient method and coordinate system-based single-objective path planning model are developed to solve the UAV’s flight path. It has the shortest trajectory, minimum threat, flight altitude, flight angle, and obstacles as constraints. In the algorithm design, quasi-opposition-based learning, real-time boundary processing, and enhanced search mechanisms are introduced into the basic whale optimization algorithm. This increases population diversity and improves the algorithm’s global optimization ability and convergence performance. Through theoretical analysis, it has been proven that the time complexity of the improved algorithm is the same as the basic whale optimization algorithm, and the convergence of the improved algorithm has been theoretically analyzed. The multi-dimensional and multi-algorithm extreme value optimization of the CEC2017 test function, as well as the optimization solution to the three-dimensional path planning problem for meteorological detection of the unmanned aerial vehicle in four complex scenarios, have been carried out. The experimental results clearly show that, considering both the maneuverability and turning performance of the UAV, the planned path can safely and efficiently avoid hazards. Compared with other algorithms, in complex environments, the path planning scheme obtained from the path planning model can, to some extent, alleviate the impact of increasing the number of trajectory points on solving performance. The coverage of the path planning method for the improved algorithm is 100%, and the success rate of path planning is increased by about 50%. The improved algorithm has excellent optimization ability and generates a higher-quality path.

Human Action Recognition (HAR) based on skeleton sequences has attracted much attention due to the robustness and background insensitivity of skeletal data. The convolutional neural network (CNN) for spatio-temporal representation learning has been widely utilized for skeleton-based HAR. However, the long-term spatio-temporal modeling and action category-specific feature attention have not been fully exploited. In order to explore the current potential of CNNs for skeleton-based HAR, a novel CNN architecture with temporal-channel attention and convolution fusion is proposed. Specially, the network architecture is composed of two novel modules, the Temporal-Channels Attention Module (TCA) and Multiscale Temporal Convolution Fusion module (MTCF). TCA module is designed to generate a temporal-channel attention matrix for different visual channels and temporal features, motivating the CNN to focus on the critical category-associated feature representation learning. Along the channels, MTCF module adapts the grouped residual connections to flexibly extend the convolutional temporal receptive field, without introducing additional parameters. By reverse stacking, MTCF module creates a bidirectional information interaction among inter-channels, compensating for the receptive field and information imbalance between subgroups from different branches. The proposed method was evaluated on three benchmark datasets, including NTU RGB-D, NTU RGB-D120 and FineGYM. The results show that the proposed TCA-MTCF method improves the CNNs’ ability to model long-term temporal features of skeleton sequences, achieving the state-of-the-art performance for HAR.

In order to solve the problem of limited driving speed of piezoelectric actuators when the piezoelectric components are small and the driving frequency is low. This paper proposes a bi-directional piezoelectric stick-slip actuator, which mainly utilizes the combination of flexible drive foot and lever amplification mechanism to realize the displacement amplification function, thus improving the driving performance. Firstly describe the structure and working principle of the proposed actuator, and carries out theoretical analysis and mathematical modeling of the flexible stator. Then the driving process of the proposed actuator is simulated and analyzed with COMSOL Multiphysics simulation software, and the feasibility of the driver structure is preliminarily verified. Finally, establish an experimental system to test the basic characteristics of the prototype. The experimental results confirm that under the piezoelectric stacks elongation of $5\mu $ m, the driving frequency of 610Hz, the maximum drive speed of the actuator is 11.21mm/s. The maximum load capacity is 0.78N and the forward and backward displacement deviation within 10 cycles is $0.99\mu $ m under the locking displacement of $10\mu $ m. The proposed stick-slip actuator is capable of realizing a large driving speed when the elongations of the piezoelectric stacks are small and the driving frequency is low.

This study introduces a novel approach to improve crop classification accuracy in airborne synthetic aperture radar (SAR) time-series imagery, focusing on overcoming the challenges posed by the incidence angle effect. The approach aims to innovate the integration of transfer learning and variational mode decomposition techniques. Transfer learning effectively addresses disparities in data distribution caused by varying incidence angles encountered in airborne SAR. Variational mode decomposition extracts robust temporal features, significantly reducing sensitivity to incidence angle variations. The approach is further enhanced by incorporating incidence angle information into the transfer learning model's training phase. The experimental results demonstrate the effectiveness of the method, which, under comparable sample conditions, achieves a remarkable improvement in accuracy (Kappa +25.05%) compared with the conventional methods. This improvement is particularly notable for crops, such as oats and soybeans, which are considerably influenced by the incidence angle effect, with Kappa increases of 27.92% and 39.30%, respectively. This study not only develops an effective strategy for crop classification in the context of airborne SAR imagery but also provides references for the effective use of new technologies from various fields in the field of remote sensing application.

Given the context of the challenge of exponentially increasing data traffic on communication networks brought by the 5G era, this paper focuses on how to apply deep reinforcement learning (DRL) techniques to solve the problem of optimizing the energy efficiency of D2D communications in heterogeneous cellular network environments. We propose a joint resource allocation scheme based on multi-intelligent deep reinforcement learning, which enables D2D devices to intelligently switch between license-free and optimal license bands and adjust the transmit power in real time to maximize the energy efficiency improvement. In this work, a multi-intelligent deep reinforcement learning framework is designed to enable D2D users in heterogeneous networks to make collaborative decisions and dynamically adjust their communication strategies according to real-time network status and environmental changes. In this paper, a deep Q-network model with a graph attention network (GAT) as the core structure is constructed; this model can cope with the complexity and diversity of network states and learn and execute optimal resource allocation strategies. In this paper, we propose a targeted loss function design that balances the optimization goal of D2D communication energy efficiency with network stability and long-term gains. Through rigorous simulation experiments, this paper verifies that a DRL-based approach can significantly improve the energy efficiency of D2D communications in heterogeneous cellular networks in real-world scenarios while ensuring the stability of the quality of service (in terms of, e.g., rate, delay, and resource utilization).

Deep chlorophyll maximum (DCM) is a common oceanographic phenomenon characterized by a significant peak in chlorophyll concentration at a specific depth below the ocean surface. DCM formation is closely related to factors, such as light availability, nutrient distribution, and ocean circulation, making it an important indicator for studying marine ecosystems and their changes. This study aims to estimate subsurface chlorophyll concentrations in the Mediterranean region using an improved multilayer perceptron model, bridging the gap between sparse observation data and dense sea surface data. We utilize Biogeochemical Argo and satellite data, including longitude, latitude, sea surface temperature, surface chlorophyll concentration, and month, as inputs to the model to estimate subsurface chlorophyll concentrations from 1 to 300 m depth. Through fitting and analyzing chlorophyll concentration data in the Mediterranean region, we explore DCM characteristics and their variations across different regions and seasons. The results indicate that the IMLP model performs excellently in estimating subsurface chlorophyll concentrations and effectively captures DCM features in various regions and seasons. By comparing the model estimations with observation data, we reveal patterns in DCM characteristics in the Mediterranean region, providing valuable data support for further research into marine ecosystems.

The prediction of the remaining useful life (RUL) of bearings is of great significance for reducing cost and increasing efficiency of mechanical equipment and ensuring healthy operation. Most of the traditional RUL prediction methods based on deep learning only analyze single sensor data, which makes the prediction accuracy fluctuate greatly and the reliability is low, and the contributions of data collected by different sensors to RUL prediction are inconsistent, which makes the low utilization of data lead to poor prediction results. To solve the above problems, a new RUL prediction framework with convolutional attention mechanism and temporal convolutional network (CAMTCN) is proposed. Firstly, the vibration signals collected by multi-sensor are normalized and processed directly as the input of the network. Secondly, an Efficient Adaptive Shrinkage (EAS) model is designed to adaptively eliminate noise interference and improve the accuracy of RUL prediction. Finally, we use the attention mechanism to design the convolutional attention subnet and combine it with the TCN to obtain the contribution degree of bearing degradation features collected by different sensors. Adaptive weight allocation captures more important degradation features to achieve end-to-end RUL prediction. Experiments were conducted on IEEE PHM 2012 dataset and XJTU-SY dataset respectively, and compared with some advanced in-depth learning methods. The experimental results show that this method can predict RUL more accurately on different bearing data sets, and has good generalization.

In an actual engineering environment, some rotating machines are usually in normal operation, but their time in a fault state is very short, which leads to a serious imbalance in the fault diagnosis datasets for rotating machinery, and gives the traditional network model the shortcomings of poor stability and low accuracy in practical engineering applications. To solve this problem, we propose a fault diagnosis method based on the combination of a new Dual-stage Attention-based Recurrent Neural Network (DA-RNN) and depth residual dispersion self-calibration convolution network (SC-ResNeSt). Firstly, a novel DA-RNN network with a gated cycle unit (GRU) as a coding-decoding unit was designed, and the network was used to predict and expand the scarce fault signals. Secondly, to make full use of the time domain information of vibration signals, a new image coding method, namely, Gram Angle Product Field (GAPF), was proposed. Then, because the traditional convolution layer lacks a dynamic receptive field to extract more representative features, self-calibrated convolution modules were introduced on the basis of the distraction network (ResNeSt), and a new network model, SC-ResNeSt, was established. Finally, the expanded vibration signal is converted into GAPF, which is used as the input for the SC-ResNeSt network to classify the fault types. To check the performance of the model, the Case Western Reserve University rolling bearing dataset and planetary gearbox dataset were used for testing. Ultimately, good results were obtained in a prediction experiment for bearing fault samples and a subsequent fault diagnosis experiment for bearings and gears, which verified the feasibility and practicability of the model.

Recent developments in brain-machine interface technology have rendered seizure prediction possible. However, the transmission of a large volume of electrophysiological signals between sensors and processing apparatuses and the related computation become two major bottlenecks for seizure prediction systems due to the constrained bandwidth and limited computational resources, especially for power-critical wearable and implantable medical devices. Although many data compression methods can be adopted to compress the signals to reduce communication bandwidth requirement, they require complex compression and reconstruction procedures before the signal can be used for seizure prediction. In this paper, we propose $\text{C}^{{2}}$ SP-Net, a framework to jointly solve compression, prediction, and reconstruction without extra computation overhead. The framework consists of a plug-and-play in-sensor compression matrix to reduce transmission bandwidth requirements. The compressed signal can be utilized for seizure prediction without additional reconstruction steps. Reconstruction of the original signal can also be carried out in high fidelity. Compression and classification overhead from the energy consumption perspective, prediction accuracy, sensitivity, false prediction rate, and reconstruction quality of the proposed framework are evaluated using various compression ratios. The experimental results illustrate that our proposed framework is energy efficient and outperforms the competitive state-of-the-art baselines by a large margin in prediction accuracy. In particular, our proposed method produces an average loss of 0.6% in prediction accuracy with a compression ratio ranging from 1/2 to 1/16.

Machine learning on electromyography (EMG) has recently achieved remarkable success on various tasks, while such success relies heavily on the assumption that the training and future data must be of the same data distribution. However, this assumption may not hold in many real-world applications. Model calibration is required via data re-collection and label annotation, which is generally very expensive and time-consuming. To address this issue, transfer learning (TL), which aims to improve target learners’ performance by transferring knowledge from related source domains, is emerging as a new paradigm to reduce the amount of calibration effort. This survey assesses the eligibility of more than fifty published peer-reviewed representative transfer learning approaches for EMG applications. Unlike previous surveys on purely transfer learning or EMG-based machine learning, this survey aims to provide insight into the biological foundations of existing transfer learning methods on EMG-related analysis. Specifically, we first introduce the muscles’ physiological structure, the EMG generating mechanism, and the recording of EMG to provide biological insights behind existing transfer learning approaches. Further, we categorize existing research endeavors into data based, model based, training scheme based, and adversarial based. This survey systematically summarizes and categorizes existing transfer learning approaches for EMG related machine learning applications. In addition, we discuss possible drawbacks of existing works and point out the future direction of better EMG transfer learning algorithms to enhance practicality for real-world applications.

Grid-forming (GFM) control based high-voltage DC (HVDC) systems and renewable energy sources (RESs) provide support for enhancing the stability of power systems. However, the interaction and coordination of frequency support between the GFM-based modular multilevel converter based HVDC (MMC-HVDC) and grid-following (GFL) based RESs or GFM-based RESs have not been fully investigated, which are examined in this study. First, the detailed AC- and DC-side impedances of GFM-based MMC-HVDC are analyzed. The impedance characteristics of GFL- and GFM-based wind turbines are next analyzed. Then, the influences of GFL- and GFM-based wind farms (WFs) on the DC- and AC-side stabilities of WF-integrated MMC-HVDC systems are compared and evaluated. The results show that the GFM-based wind turbine performs better than the GFL-based wind turbine. Accordingly, to support a receiving-end AC system, the corresponding frequency supporting strategies are proposed based on the GFM control for WF-integrated MMC-HVDC systems. The GFM-based WF outperforms the GFL-based WF in terms of stability and response time. Simulations in PSCAD/EMTDC demonstrate the DC- and AC-side stability issues and seamless grid support from the RESs, i.e., WFs, to the receiving-end AC system.

To utilize the load-bearing potential of permanent magnetic levitation (PML) structures and improve the levitation efficiency of rare-earth permanent magnetic materials, firstly, the number of magnetic blocks per unit wavelength and the correction factor of magnetic induction strength are introduced to establish an analytical model of vertical levitation force for the PML system based on Halbach arrays, which can effectively make up for the lack of precision caused by the production and installation of permanent magnetic arrays. Secondly, a method to optimize the dimensional parameters of the magnetic track structure combining quantitative analysis and multi-parameter optimization is proposed. Thirdly, a new permanent magnetic levitation structure with high levitation efficiency is designed. The experimental results show that the maximum error between the analytical model and the finite element calculation results is about 5.50% and the minimum error is 0%. The proposed magnetic circuit optimization structure improves the levitation efficiency by 4.60% and the ratio of levitation force to lateral deflection force by 2.48% compared with the “red track”. This study provides good reference for the design of permanent magnet magnetic levitation systems.

Traffic sign recognition and detection is a key technology in automatic vehicle driving and driver assistance systems. However, existing traffic sign recognition algorithms suffer from problems such as large model size, complex computation, high computational cost, which make it difficult to achieve an effective balance between detection speed and detection accuracy. This paper proposed an improved lightweight recognition algorithm, which is based on YOLOv5. This algorithm replaces the convolutional structure in the original YOLOv5 neck network with Ghost Module and C3Ghost Module, thereby reducing redundant features in the feature fusion process, lowering computational cost and the number of parameters. The structure of the PAN network was improved and the hybrid attention mechanism module CBAM was introduced to capture key information in traffic signs. Cross-layer connections were added to shorten the path of information transfer in feature pyramid network, which fused more features and improved the network feature recognition accuracy. In addition, the EIoU_Loss function was adopted as the bounding box regression loss function to improve the localization accuracy of the algorithm. The performance of the improved algorithm was also verified on the Chinese traffic sign dataset. Experimental results showed that the improved algorithm’s detection accuracy was enhanced by 1.2%, while mAP@0.5 and mAP@0.5:0.95 were enhanced by 1.5% and 3.4% respectively over the existing YOLOv5 algorithm, and the overall parameter numbers and computational cost of the model were reduced by 14.5% and 16%. The proposed algorithm performs better than the current mainstream detection algorithms, has higher recognition accuracy in multiple environments, and meets the demand for real-time traffic sign recognition.

Aiming at the shortcomings of the path generated by the artificial potential field (APF) method, such as local minimum, target unreachability, and low path smoothness, an improved artificial potential field method is proposed. First, to reduce the collision risk and planning difficulty, based on known environmental information such as the location of obstacles and targets, the area with fewer obstacles is selected as the priority area for path planning. Second, to improve the path smoothness and reduce the computation amount, an adaptive step-size adjustment method based on the distance and angle relationship with obstacles within the prediction range is proposed. Third, in view of the effect on each other between obstacle, local minimum, and unsmooth path, a multi-target model considering the size and influence range of obstacles and an improved potential field function are proposed on the basis of the identified planning priority area. Finally, in order that the path is smooth enough to be tracked by autonomous mobile robots, a safe driving corridor without collision with obstacles is constructed on the planned path, and a trajectory fully constrained to the safe driving corridor is generated using the quadratic programming method. The simulation comparison experiments are carried out on matlab simulation software and the smoothness of IAPF is improved by an average of 97.3% as compared to traditional APF and 45.19% as compared to DWA. The sum of the proposed IAPF path planning and optimization time is improved by 45.1% on average compared to DWA path planning time.

This paper proposes an asymmetric encryption and compression method based on three-dimensional Tent-Cubic-ICMIC map (3D-TCICM) and elliptic curve encryption (ECC) for 3D remote sensing point cloud. Due to the large size, rich data and high security of remote sensing point cloud, an encryption and compression method with high computing efficiency and high security is required. The point cloud is first divided into blocks, then compressed by point cloud library (PCL) using octree, and then the resulting data stream is encrypted. During encryption, a chaotic system 3D-TCICM with excellent chaotic behavior is proposed. Its initial values are generated by point cloud data and ECC algorithm. The sender and receiver only hold the public key or private key respectively, which further improves the security of the algorithm. The encryption factors are also generated for subsequent encryption operations, which improves the plaintext correlation. The encryption algorithm includes innovative scrambling algorithm between multiple blocks (MBS), helix diffusion in GF($2^{8}$) field and multiple S-boxes substitution, in which different S-boxes are provided for each data block. Simulation results indicate that the proposed scheme surpasses existing algorithms in terms of security, randomness, and plaintext correlation, all while maintaining a lower computational complexity. The algorithm exhibits robustness against various attacks such as differential, statistical, chosen-plaintext, known-plaintext, chosen-ciphertext, noise, and cropping attacks.

A subsurface chlorophyll maximum is an important ecological feature of planktonic ecosystems. Although the vertical profiles can be determined through the implementation of biogeochemical (BGC)-Argo buoy, this method is not compatible with the ocean observation requirements of high-resolution spatiotemporal measurements. Here, we demonstrate that deep learning can proficiently fill in these observational gaps when combined with sea surface data from ocean color radiometry. First, the sparse vertical profile data of BGC-Argo is fused with sea surface data to construct the benchmark dataset for deep learning. Second, encouraged by the idea of dense numerical representations, the comprehensive model combined with coupled embedding and bidirectional gated recurrent unit is proposed to inverse the vertical profile with BGC-Argo and satellite data. Then, the in-depth spatiotemporal analysis of the subsurface chlorophyll maxima phenomenon is performed by the parametric equation method and deep learning method as well. Finally, extensive experiments in the Northwest Pacific were conducted to demonstrate the effectiveness of the proposed methodology. The impressive results indicate that the proposed method can compensate for the lack of sparse in situ observations of chlorophyll concentration, the determination coefficient is increased by more than 20%. This study is of great significance to marine ecology and provides important insight into artificial intelligence in the study of subsurface oceanic phenomena.

During the construction of water conservancy and hydropower projects, the time series of safety monitoring data for slope engineering is often short. When establishing a slope safety monitoring model, it is necessary to study the modelling method under small sample conditions and closely examine the space–time information in the monitoring data for the limited length of each measuring point. To this end, this paper initially proposes using the Ward clustering method to spatially cluster anchor cable prestress measuring points in different parts of a slope; then, according to the clustering results, a safety monitoring model for slope anchor cable prestress is established for each type of measuring point based on a Bayesian panel vector autoregressive (BPVAR) model with highly accurate small sample modelling. Finally, slope anchor cable prestress monitoring data in China are taken as examples for verification analysis. The analysis results show that the prestressed measuring points of the slope anchor cable for this project are divided into four categories: excavation causes the prestress of some slope measuring points to continue to increase, and the tension or compression of the structural plane leads to an increase or decrease in prestress. The multiple correlation coefficients of the training set and test set data of the BPVAR model are all above 0.80, and the prediction error of the validation set is less than that of the vector autoregressive (VAR) model, the autoregressive moving average (ARMA) model and the long short-term memory (LSTM) model. The measured prestress values are all within the 95% confidence interval, which provides a reference for safety state identification in slope engineering.

Laser-based illuminant has been a topic of increasing interest in recent years and both of color quality and blue light hazards are the key issues that need to be addressed. In this paper, the performances of color quality and blue light hazard were investigated by comparing 465 nm laser-excited phosphor materials and traditional 452 nm laser-excited phosphor materials respectively. Phosphor-converted laser-based illuminant (pc-LBI) consisting of 465-nm and 635-nm LDs (LBI-465) has a color rendering index (CRI) of 88.7, which was 8.2% higher than that of pc-LBI consisting of 452-nm and 635-nm LDs (LBI-452). The correlated color temperature (CCT) of white light of both pc-LBIs ranges from 2200 K to 5000 K with the adjustment of red lasers. The blue light hazard efficiency of radiation (BLHER) of LBI-465 is below 0.15, which is 19% lower than that of LBI-452. The results showed that high color quality and low blue light hazard were simultaneously achieved by LBI-465. In addition, the above experimental results are in good agreement with the simulated results. Based on these above results, pc-LBI is believed to render laser-based white light sources with good color performance and boost the growth of solid-state lighting.

Most of the existing steganalysis methods are designed for specific steganography methods in low-bit-rate compressed speech stream and lack of generalization ability. In practical applications, the steganography methods in compressed speech are various and cannot be predicted in advance. We can only employ numerous possible steganalysis method to detect, which is laborious and time-consuming, and cannot achieve real-time detection. Therefore, it is necessary to develop a general steganalysis method that can detect multiple steganography methods simultaneously for compressed speech. To this end, a steganalysis method based on global and local correlation mining is proposed in this paper. Firstly, a codeword distributed embedding module is introduced to transform the compressed codewords into a compact feature representation. Then, global-guided correlation mining module and local-guided correlation mining module are used to extract the correlation change before and after steganography in the view of global and local. Finally, the detection results can be obtained by the full connection layers. Experimental results show that the proposed method can reach a better detection performance than the existing steganalysis methods at different embedding rates and speech lengths.

Impervious surface area (ISA) mapping at the global scale has entered a new era. Currently, the number of high-resolution global ISA products is gradually increasing; however, a literature review that systematically investigates these ISA products is still lacking, which limits the application of these products. Thus, we provide a comprehensive analysis of the existing high-resolution global ISA products, concentrating on the aspects of the data sources, training samples, features, and methods. Moreover, we evaluate these products at multitemporal and multispatial scales, using a series of independent test samples. The results demonstrate that the multitemporal accuracy of the ISA products presents an increasing trend, due to the increase of the available sensors. Among the continuous time-series products [e.g., the updated new global impervious surface area (GISA 2.0), the global impervious surface area (GISA), global annual urban dynamics, global human settlement layer, and global artificial impervious areas], the accuracy of the GISA 2.0 outperforms the others at global, continental, and regional scales. However, the mapping performance of these products in small towns and arid and rural regions needs to be enhanced. In particular, we focus on the spatio-temporal disagreement of the ISA products. We show that the high disagreement regions are predominantly concentrated in eastern Asia, western Europe, and eastern North America. In addition, the high disagreement regions are characterized by low ISA density, high vegetation coverage, and high albedo bare ground coverage. Additionally, this article concludes with some remarks about the future directions of global ISA mapping.

When used for object tracking, the discriminative correlation filter (DCF) is effective, but its performance is often burdened by undesirable boundary effects. Meanwhile, when there is too much background information in training samples of the DCF, it will be easier to learn the area deviating from the tracking object. Further, illumination variation, partial/full occlusion, and appearance variations, render the response map aberrance of the correlation filter (CF) more prone to occur. To overcome these problems, an object tracking model based on a time-varying Spatio-temporal regularized correlation filter with aberrance repression is proposed in this paper. Firstly, by adding a regularized term to the traditional CFs to limit the change rate of the response map generated in the object detection phase, the proposed tracker can obviously repress the aberrance of the response maps; secondly, by adjusting the filter to the object regions suitable for tracking with high confidence scores with a time-varying spatial reliability map, the proposed tracker effectively overcomes the adverse effects caused by the boundary effect; and finally, by introducing a temporal regularized term, the proposed tracker also has superior tracking ability for the partial occluded objects and those with large appearance variations. Significant experiments on the OTB100, VOT2016, TC128, and UAV 123 datasets have revealed that the performance thereof outperformed many state-of-the-art trackers based on DCF and deep-based frameworks in terms of tracking accuracy, tracking success rate, and A-R rank, etc.

Currently, steganography based on compressed speech streams is gathering more and more attention. Meanwhile, it poses a huge threat to cyber security. As a counter technique, steganalysis can detect whether an illegal secret message is embedded in a compressed speech. To further improve the detection performance of current methods, a novel steganalysis method based on codeword association rule mining (CARM) is proposed in this paper. Firstly, we analyzed the spatiotemporal relationships between codewords in compressed speech. Secondly, the steganography-sensitivity codeword association rule base in training set was built based on the confidence change of codeword association rules before and after steganography. Thirdly, the steganography characteristic index and the corresponding dynamic partition threshold in validation set were computed to determine whether the compressed speech segment contains covert communication or not. Finally, comprehensive experiments were conducted to evaluate the performance of the proposed CARM steganalysis method under various conditions, including different association rule patterns, whether to use dynamic partition threshold, different embedding rates, different speech lengths, etc. The experimental results verify that CARM can achieve better performance than the comparison methods. In addition, the detection accuracy of CARM method can be improved significantly by using dynamic partition threshold at low embedding rates.

In traditional CSMA/CA-based MAC designs supporting successive interference cancellation (SIC), channel utilization is inefficient, because the inherited CSMA/CA-style contention approach often forces the whole channel to remain idle and the early released channel is not utilized. In this paper, we propose Multi-CQ, which is the first software-defined networking (SDN)-based MAC design for wireless LANs that supports SIC and significantly improves the channel utilization. Multi-CQ, adopting SDN’s functional separation idea and OFDMA, makes contention and data transmission be executed over two subchannels independently and concurrently, where the superposition coding are used for decoding combined signals, and multiple CQs (contention queues) are introduced to coordinate the concurrency. This concurrent execution greatly reduces the waste in channel contention. Next, adopting SDN’s central control idea, Multi-CQ controller selects nodes who can finish their frame transmissions in almost equal time, thereby improving the channel utilization. Finally, we develop a theoretical model to optimize bandwidth allocation for channel contention and data transmission. Extensive simulations verify the effectiveness of our design and the accuracy of our theoretical model. This study is also helpful to better design wireless MAC protocols supporting other advanced functionalities such as MU-MIMO.

A build-in data inversion method to retrieve aerosol size distributions based on the principle of particle electrical mobility has been introduced, whose required computational effort is low for a portable ultrafine particle sizer (PUPS). The PUPS is cost-effective for the measurements of the fine and ultrafine particles in polluted environments near-source (e.g., cities with high traffic density, freeways, airports or stationary combustion sources). The particle sizer is mainly composed of a unipolar charger, a plate differential mobility analyzer (PDMA), and a Faraday cup electrometer (FCE). The classification efficiencies of the PDMA are strongly dependent on the charging distribution of the unipolar charger, in which multiple charging is more significant than that for a bipolar charger. To reduce the excessive overlap in the kernel function caused by multiple charging, a guidance method for selecting the classification voltages and operating parameters of the PDMA is proposed with the help of MATLAB. Subsequently, a combination of the nonnegative least squares (NNLS) algorithm and post facto smoothing method has been employed to derive the discretized solution of the Fredholm integral equation of the first kind. In the end, the accuracy and stability of the proposed approach are tested under a range of particle size distribution scenarios. The synthesized data results show that for the unimodal aerosol distribution, almost all of the relative errors are less than 20 %, regardless of nonideal operating conditions. The inversion algorithm can be run on Cortex-M3, an ARM embedded chip for low-cost and low-power consumption applications. When a reasonable error range can be permitted, the inversion algorithm can meet the requirements of the PUPS.

Bad weather, such as snowfall, can seriously decrease the quality of images and pose great challenges to computer vision algorithms. In view of the negative effect of snowfall, this paper presents a single-image snow removal method based on a generative adversarial network (GAN). Unlike previous GANs, our GAN includes an attention mechanism in the generator component. By injecting attention information, the network can pay increased attention to areas covered by snow and improve its capability to perform local repairs. At the same time, we improve the traditional U-Net network by combining it with the residual network to enhance the effect of the model when removing snowflakes from a single image. Our experiments on both synthetic and real-word images show that our method produces better results than those of other state-of-the-art methods.

The alpha approximation method is known to be effective and simple for soil moisture retrieval from time series of synthetic aperture radar data. However, its accuracy is usually degraded by the scattering from vegetation, and it entails working with an underdetermined linear system when solving the unknown surface parameters. In this work, we study how the availability of fully polarimetric data and a diversity in incidence angles can help this method for soil moisture estimation. Results are obtained using data from the Soil Moisture Active Passive Validation Experiment 2012 campaign acquired by an air-borne L-band radar system. The assessment of the performance is based on in situ measurements over agricultural fields corresponding to five different crop types: bean, soybean, canola, corn, and wheat. The validation shows that, compared with the original method, the retrieval accuracy can be improved when the polarimetric decomposition is included in the approach. The combination of polarimetric decomposition and multi-incidence observations of enriched data provides the best performance, with a decrease in the final root-mean-square error between 0.4% and 5% with respect to single-pol and single-incidence data. Compared with HH, the results obtained for VV data present a higher accuracy for the overall crop types. The most noticeable improvement is achieved for corn, soybean and wheat, demonstrating the contribution of this extension of the original approach.

Assessing cross-space risk of cyber-physical distribution system under integrated attack is investigated in this paper. Firstly, a hierarchical structure of cyber-physical distribution network according to IEC 61850 is established and a deliberate attack scenario with limited adversarial knowledge and stealth requirement is developed based on a general linear model for state estimation. Then, we formulate two optimization problems to describe the attack implementation and propagation process and obtain the likelihood of attacks including a robust solution and a risk solution in a fuzzy Bayesian network (BN). On this basis, a physical impact metric is defined as the integrated deviation of system states and measurements. Thus, the cross-space risk assessment can be performed. Finally, the simulation results of case studies demonstrate that the proposed method is effective and provides a broad and clear view of cyber-physical distribution system security situation.

Flexible job-shop scheduling problem (FJSP) is one of the most important optimization problem in manufacturing domain. With the development of manufacturing industry, the dimensionality of FJSP increases and its search space expands exponentially. It is hard to obtain a desired scheduling solution in an acceptable time using the traditional evolutionary algorithms. In this paper, we propose a hybrid distributed evolutionary model for large scale flexible job-shop scheduling problem (LSFJSP). The model is composed of two layers: division layer and coevolution layer. In the division layer, a grading mechanism and a Bayesian-grouping method are developed to decompose population and dimension respectively to obtain several subpopulation. In the coevolution layer, an individual migration mechanism and a probability selection mechanism for reference vector are used to achieve the coevolution among the subpopulations. Three typical evolutionary algorithms are integrated in the proposed model to test its superiority. The experimental results on the large scale instances of LSFJSP show that the evolutionary algorithms integrated in the proposed model have better optimization results and higher computational efficiency in comparison with the corresponding evolutionary algorithms in other distributed models.

With the development of specialization, coordination and intelligence in the manufacturing service process, the issue of how to quickly extract potential resources or capabilities for distributed manufacturing service requirements, and how to carry out resource matching for manufacturing service requirements with correlated mapping characteristics, have become the critical issues to be addressed in the cloud manufacturing environment. Through the combination of the characteristics of relevance, synergy and diversity of manufacturing service tasks on the intelligent cloud platform, a matching decision method for manufacturing service resources is proposed in this paper based on multidimensional information fusion. On the basis of integrating multidimensional information data in cloud manufacturing resource, the information entropy and rough set theory are applied to classify the importance of manufacturing service tasks, while the matching capability are analyzed by using a hybrid collaborative filtering (HCF) algorithm. Then, the information of function attribute, reliability and preference is employed to match and push manufacturing service resources or capabilities actively, so as to realize the matching decision of manufacturing service resources with precise quality, stable service and maximum efficiency. At last, a case study of resources matching decision for body & chassis manufacturing service in a new energy automobile enterprise is presented, in which the experimental results show that the proposed approach is more accuracy and effective compared with other different recommendation algorithms.

In this paper, we propose a specular highlight image enhancement algorithm based on a dark channel prior to solve the problem of information loss in specular highlight images in real scenes. First, the algorithm is based on the dark channel prior algorithm, and a moving window minimum filter is used to estimate the global illumination component. Then, a weighted function based on the local pixel color difference is introduced to solve the halo artifacts in the image. Then, the improved guided filter algorithm is used to optimize the transmittance, which improves the computational efficiency of the algorithm. Finally, the brightness of the image is adjusted by the CLAHE algorithm to enhance the local details of the image. Experimental results show that this method can effectively enhance the information in specular highlight images, and its processing results are obviously better than those of other algorithms.

Specular highlights often exist in real-world images due to the material property of objects and the capturing environments. High light in real-life scenes can reduce image quality and cause loss of image information, to overcome this, we propose an effective specular reflection image enhancement algorithm. First, the scene depth map of specular reflection image was obtained by the color attenuation prior method; then, two adaptive adjustment factors were developed to optimize the transmission map for specular reflection images and $L_{0}$ gradient minimization filter was used to eliminate halo artifacts. Finally, the nonuniform illumination compensation strategy based on the YUV color space was used to expand the dynamic range of the image and increase the saturation of the image. Experimental results via subjective and objective evaluations demonstrate that the proposed method can provide better visual quality than other methods and can effectively enhance the information in the specular reflection image.

Video desnowing has become a challenging research topic in computer vision in recent years. Existing methods cannot remove most of the snow in heavy snow scenes and will cause the deformation of moving objects when used for snowy videos that include moving objects. These methods have poor generalizability, exhibiting poor performance when removing snow from videos with different resolutions. In this paper, we propose a new video snow removal method based on self-adaptation snow detection and a patch-based Gaussian mixture model (VSRSG). First, an optical flow estimation method and a support vector machine (SVM) are used to detect snowflakes, and a self-adaptation threshold process is used to remove dense snowflakes in the snowflake detection map to obtain a sparse snowflake detection map. Then, a patch-based Gaussian mixture model (PBGMM), which can remove moving objects and both sparse and dense snowflakes from videos and restore a clear video background, is applied for background modeling. A Markov random field (MRF) and self-adaptation threshold processing are used to extract sparse snowflakes and moving objects and combine them with the background to form an input video without dense snowflakes. Finally, a similar block matching method is employed to fill in the detected snowflake pixels with the information from adjacent frames to remove the sparse snowflakes in the near range. This method can also remove snowflakes in front of moving objects. Experiments show that the proposed method can simultaneously remove sparse snowflakes, dense snowflakes and snowflakes in front of moving objects and outperforms other state-of-the-art methods.

This paper addresses issues of brain tumor subtype classification using Magnetic Resonance Images (MRIs) from different scanner modalities like T1 weighted, T1 weighted with contrast-enhanced, T2 weighted and FLAIR images. Currently most available glioma datasets are relatively moderate in size, and often accompanied with incomplete MRIs in different modalities. To tackle the commonly encountered problems of insufficiently large brain tumor datasets and incomplete modality of image for deep learning, we propose to add augmented brain MR images to enlarge the training dataset by employing a pairwise Generative Adversarial Network (GAN) model. The pairwise GAN is able to generate synthetic MRIs across different modalities. To achieve the patient-level diagnostic result, we propose a post-processing strategy to combine the slice-level glioma subtype classification results by majority voting. A two-stage course-to-fine training strategy is proposed to learn the glioma feature using GAN-augmented MRIs followed by real MRIs. To evaluate the effectiveness of the proposed scheme, experiments have been conducted on a brain tumor dataset for classifying glioma molecular subtypes: isocitrate dehydrogenase 1 (IDH1) mutation and IDH1 wild-type. Our results on the dataset have shown good performance (with test accuracy 88.82%). Comparisons with several state-of-the-art methods are also included.

Due to the scattering and absorption of light by sandstorm particles, images taken in sandstorm environments have low contrast, color distortion and other degradation problems, which seriously affect outdoor computer vision systems. In this paper, we propose a novel method to enhance images captured in sandstorm weather conditions. First, the method builds on the blending of two images that are directly derived from the original degraded image. Second, we use multilayer decomposition technology to enhance image details and use a blue channel and white balancing technology to restore image contrast and chromaticity. Third, we associate weight maps to improve image edge contrast. Finally, the Laplacian pyramid fusion method is used to obtain the fusion results of the sandstorm-free color correction image. The experimental results demonstrate that the proposed method can effectively restore the fade characteristics of sandstorm-degraded images and improve the clarity of the images. Experimental results via subjective and objective evaluations demonstrate that the proposed method can significantly improve images captured during sandstorm weather conditions, and the results are better than other methods.

In natural language processing, semantic document exchange ensures unambiguity and shares the same meaning for documents sender and receiver cross different natural languages (e.g., English to Chinese), this difference makes the translation between natural languages becomes complex and inaccurate. This paper proposed a novel framework of Context-Free Machine Universal Grammar which consists of local mode (sender and receiver) and mediation mode (Machine Universal Language) based on the concept of collaboration, the framework improves semantic unambiguity and accuracy in crossing language document, meanwhile makes document computer-readable through unique ID for each word or phrase. More importantly, inspired by grammatical case in linguistics, a novel Machine Universal Grammar provides a universal grammar that accepts all coming languages and improves semantic accuracy in natural language processing.

In recent years, trackers based on correlation filters have attracted more and more attention due to the impressive tracking accuracy and real-time performance. However, in real scenarios, the tracking results are often been interfered with by the occlusion, illumination variation, appearance variation and background clutter. In order to find a tracker with better tracking performances, this paper proposed a multi-information fusion correlation filter tracker, which uses channel and spatial reliabilities and time regularization information on samples for filter training, and which not only extends the target search areas but also has a stronger ability to track the targets with significant appearance variations. Thus, results from extensive experiments conducted on OTB100, VOT2016, TC128, and UAV123 data sets show that our tracker with only directional gradient histogram (HOG) and color name (CN) features, performs favorably against the state-of-the-art trackers in terms of tracking precision, tracking success rate, tracking accuracy, and A-R rank.

Diffusion-weighted magnetic resonance imaging (DWI) is sensitive to acute ischemic stroke and is a common diagnostic method for the stroke. However, the diagnostic result relies on the visual observation of neurologists which may vary from doctor to doctor under different circumstance. And manual segmentation is often a time-consuming and subjective process. The time from onset to thrombus removal has a significant impact on the prognosis of patients with acute ischemic stroke. The shorter the time, the better the prognosis. For this purpose we present a novel framework to quickly and automatically segment the ischemic stroke lesions on DWI. We mainly have three contributions: firstly, we design a detection and segmentation network (DSN) to solve the two kinds of data imbalance; secondly, we propose a triple-branch DSN architecture, used for extracting different plane feature respectively; thirdly, we propose a multi-plane fusion network (MPFN), which aims to make final prediction more accurate. Extensive experiments on ISLES2015 SSIS DWI sequence dataset demonstrate the superiority of our proposed segmentation method. The dice reached 62.2% and the sensitivity reached 71.7%.

The interaction between a moving submerged body and a homogeneous/stratified fluid generates a Bernoulli hump, a Kelvin wake, and an internal wake on the water surface. The height distribution of wakes is directly related to the location and motion state of the submerged body. Thus, it is possible to retrieve kinematics information of a submarine from the images of wakes obtained by photoelectric detection equipment. In order to extract the submarine's location and velocity information effectively from the wake photoelectric images, we investigate the relationship between the motion state of moving submerged bodies and wakes in this paper. Bernoulli hump, Kelvin wake, and internal wake models are established based on potential flow theory and thin-ship approximation. We analyzed the wave crestline pattern and the energy characteristics of Kelvin wake components (divergent and transverse waves) and internal wakes at different velocities and diving depths. Finally, we propose a method for estimating the velocity of a submarine based on the wake wavelength and the diving depth inversion method based on the Fourier power spectrum of the Kelvin wake. The results obtained prove the feasibility of using photoelectric equipment to obtain wake images for use in analyzing the kinematics state of submarines, which are of guiding significance for detection and information processing in real-scale submarines.

The single-axis maglev system is increasingly popular as undergraduate project kits over the recent years. Though it is simple and instructive, the large current in the electromagnet leads to the overheating problem. In order to enhance the energy-saving performance as well as the controller performance, this work compares three geometric modifications on the iron core, the upper permanent magnet and the floating permanent magnet for the maglev system. Four target cases are defined to incorporate the geometric modifications and are solved numerically. Moreover, the numerical solutions are carefully analyzed in terms of the zero-power force, the controller-gain requirement and the saturation current. Consequently, two approaches, i.e., extending the iron core and enlarging the floating magnet, can improve both the zero-power force and the controller-gain requirement and are highly recommended for the zero-power maglev system. On the contrary, though the upper magnet can improve the zero-power force, it significantly raises the controller-gain requirement and accelerates the saturation of the iron core.

Profit allocation plays an important role in the decision-making field. In this paper, we study an allocation method on restricted coalition cooperation with intuitionistic fuzzy coalitions, in which the partners have some hesitation degree and different risk preferences when they participate in a limited communication structure game. In order to sufficiently analyze the profit allocation strategy, an average tree solution (A-T solution) with Choquet integrals and hesitation degrees is studied. In particular, a simple solving method for the A-T solution is proposed by proving that the characteristic functions of the cooperative game satisfy the monotonicity condition. Using this method, the upper and lower bounds of the A-T solution can be calculated directly from the upper and lower bounds of the interval characteristic functions. This method avoids the subtraction of interval numbers. Furthermore, the properties of the A-T solution according to an axiomatic system are proved in this paper. Finally, the applicability and superiority of the proposed approach are demonstrated through comparison with other methods.

The high-voltage direct current (HVDC) grid has been recognized as an effective solution for renewable energy integration. Currently, two main development trends for HVDC grids are being studied: a DC breaker based HVDC grid and fault-blocking converter based HVDC grid. Although the former has a perfect performance for fault clearance, its development is still highly constrained by the cost and maturity of DC breakers. The latter can extinguish DC faults by the fault-blocking converters. Without using DC breakers, there is no bottleneck in its technical feasibility. Nevertheless, in fault scenarios, such types of HVDC grids will be blocked at length for air-deionization, which is its main drawback. The aim of this paper is to minimize its power interruption time, by optimizing protection coordination strategies. To cover the most complex cases, the overhead line applications, in which the reclosure actions are required to be implemented, are considered. In this paper, the protection requirements of HVDC grids are first discussed, then the benefits of fault-blocking modular multilevel converters (MMCs) and their fault features are analyzed. Based on this, a control function is designed to reduce the air-deionization time. To minimize the influence of the DC faults, a separation methodology for restarting the system is proposed. The effectiveness of the proposed protection coordination schemes is validated by PSCAD/EMTDC simulations.

Video surveillance may involve the simultaneous monitoring of a large number of areas. Real-time automatic change detection of a monitoring area (such as involving the movement of people or vehicles) can reduce risks incurred in negligent manual observation. However, the low signal-to-noise ratio (SNR) of dark environments can significantly corrupt camera images, making it difficult for machine learning surveillance systems to detect small changes in monitored images. In addition, in the absence of changes between two multitemporal monitoring images, sensor noise can lead to false alarms. The objective of this paper is to reduce the effect of sensor noise on change detection of monitored images and the run time of change detection algorithms. For these purposes, we proposed a novel multitemporal monitoring image change detection algorithm based on morphological structure filtering and normalized fusion difference image. First, the random noise in two surveillance images was removed using a multidirectional weighted multiscale series of a morphological filter. Next, two difference images were obtained by using the compression log-ratio operator and the mean ratio operator, and a fusion difference image was obtained by equal-weight fusion of the two difference images. Then, the sigmoid function was used to compress the fusion difference map to obtain a normalized fusion difference image, and a median filter was used to obtain a final difference image. Finally, the k-means clustering algorithm was utilized to obtain the change detection results. The experimental results demonstrate that the proposed method can accurately detect changes in a night monitoring scene in real time. Subjective and objective evaluation of the experimental results demonstrate that the proposed method is superior to reference algorithms in terms of change detection accuracy, time and robustness.

With the increasing cases of thyroid malignant tumors, the diagnosis of thyroid nodule has attracted more and more attention. Deep learning has achieved promising results in computer-aided diagnosis due to the advantages of obtaining high-dimensional features. In this paper, we proposed a hybrid multi-branch convolutional neural network based on feature cropping method for feature extraction and classification of thyroid nodule ultrasound images. Firstly, we designed a backbone convolutional neural network to extract shared feature maps and a classification network as global branch. Next, we added a feature cropping branch in the network to perform multi-cropping on batch feature maps, to reduce the impact on classification caused by the similarity of local features between benign and malignant thyroid nodule images. Finally, based on softmax predictions of different branch feature maps, we utilize a weighted cross-entropy loss function to train our proposed binary-classification network. Experimental results show that our proposed method has achieved 96.13% accuracy, 93.24% precision, 97.18% recall, and 95.17% F1-measure in public dataset and local dataset, outperforming other models.

The images captured in sand-dust weather have the characteristics of color deviation and low visibility, which seriously affect computer vision systems. To solve the above problems, we propose a fast and effective algorithm to enhance the images captured in sand-dust weather conditions. First, we compensate for the loss value in the blue channel. Then, white balancing technology is used to correct the color of the sand-dust-degraded image. Finally, guided image filtering is used to enhance the image contrast and edge accuracy, and an adaptive method is used to calculate the magnification factor of the detail layer to enhance the image detail information. The experimental results on a large number of sand-dust-degraded images show that the method can effectively recover the fading characteristics of sand-dust-degraded images in a short time and improve the clarity of the images. Experimental results via qualitative and quantitative evaluations demonstrate that the proposed method can significantly improve the images captured during sand-dust weather conditions, and the results are better than those of other methods.

When shooting a moving object, as the object moves too fast or the camera's exposure time is too long, smears may occur in the image, which would result in motion blur. The blind restoration of object motion blur is a challenging inversive problem. To effectively extract useful information from blurred images, this paper proposes a new method to remove motion blur, which is based on the maximum a posterior (MAP) framework. Firstly, the framework combines guided filtering and automatic GrabCut image segmentation algorithm in order to divide the image into different layers. Afterwards, it uses the image gradient to estimate the blur kernel through an alternating iterative optimization strategy. The iteratively reweighted least squares algorithm (IRLS) is used to optimize the solution of the model. Finally, we use the unsharp masking algorithm to improve the high-frequency components of the image and enhance the edge and details of the image. Therefore, the algorithm can effectively remove the blur caused by the motion of the object, suppress the noise and ringing effect, and recover a higher quality clear image, which can be demonstrated on benchmark problems.

Due to more or less deviations in the imaging system, there will be noise in the image, which makes the image segmentation inaccurate. To divide a natural image into a more accurate binary image, the target and background of the image are effectively separated to achieve a more effective segmentation result. Therefore, this paper proposes an image segmentation algorithm combining a saliency map and neutrosophic set (NS) theory. First, to overcome the problem of weak edges in the image, we highlight the details and use the guided filter to filter the various channels of the natural image. Then, the initial saliency map is generated. After the weighted superposition of the initial saliency map, the local entropy map and the gray scale map, the final saliency map can be generated using the nonlinear function, and it can effectively highlight the foreground information of the image. Second, the saliency map is transformed to the NS domain and interpreted by three subsets: true (T), indeterminate (I), and false (F). According to NS theory, the indeterminacy is reduced, and the segmentation results are finally obtained by using the method of threshold. Various experiments were done and compared with other state-of-the-art approaches. These experiments demonstrate the effect of the proposed work, which is fast and effective for de-noising.

As effective measures of energy savings and emission reduction, non-linear loads, such as variable-frequency drives, are widely used. However, the problem of harmonic pollution in power systems is becoming increasingly severe. A harmonic on-line monitoring system is required for statistically analyzing the harmonic distribution rule in a power network and for taking harmonic suppression measures. For solving the frequent topology adjustment issue in practical distribution networks, a new observability analysis method is proposed. The probability of a switch having a closed status is used to express the nodal observability degree in consideration of network topology changes. The condition number of the measurement matrix is used as the index for estimating the precision of the harmonic state. An optimal placement model of harmonic measuring devices is presented, which considers the system observability and the condition number of the measurement matrix. The genetic algorithm (GA) is improved to solve the optimal model. A simulation model is developed based on the improved IEEE33-node system. The simulation results demonstrate the validity of the optimal harmonic measuring device placement model in distribution networks.

Garbage classification has always been an important issue in environmental protection, resource recycling and social livelihood. In order to improve the efficiency of front-end garbage collection, an automatic garbage classification system is proposed based on deep learning. Firstly, the overall system of the garbage bin is designed, including the hardware structure and the mobile app. Secondly, the proposed garbage classification algorithm is based on ResNet-34 algorithm, and its network structure is further optimized by three aspects, including the multi feature fusion of input images, the feature reuse of the residual unit, and the design of a new activation function. Finally, the superiority of the proposed classification algorithm is verified with the constructed garbage data. The classification accuracy of the proposed algorithm is enhanced by 1.01%. The experimental results show that the classification accuracy is as high as 99%, the classification cycle of the system is as quick as 0.95 s.