Search

In the immediate aftermath of large-scale disasters, emergency logistics services play important roles in saving lives and reducing losses. Efficient relief logistics scheduling depends on the accurate transport time information for available routes. However, this information cannot be obtained precisely until a vehicle uses the road. Considering the correlation between information acquisition and logistics operations, this paper focuses on a multiperiod online decision-making problem to simulate the information acquiring process. This problem can be referenced for emergency resource scheduling scenarios in which previous decisions impact knowledge for future logistics plans. A multi-trip cumulative capacitated vehicle routing problem with uncertain transportation time is investigated as the basic model. The tradeoff between transportation efficiency and the unknown transport time discovery rate is considered in a multiobjective evolutionary algorithm (MOEA). A memetic algorithm (MA) and a robust optimization (RO) -MA for single-period post-disaster emergency logistics are also proposed to solve the problem for comparison. In these algorithms, evolutionary operators that benefit solution fixing and variation are proposed. In the experiments, a real-world instance is employed. A simulative experimental environment is established. Dynamic information gained within the process of logistics scheduling is highlighted via multi-period online optimization. Different scenarios corresponding to estimates in emergency situations are provided to validate the performance of the algorithms. The experimental results show that the hybrid strategy, MOEA+MA, can obtain the best result in more than half of the considered cases which demonstrates the necessary balance between obtaining information and transportation efficiency.

With the increasing capacity of wind turbines, the importance of the yaw system has gradually become more prominent. The supervisory control and data acquisition (SCADA) system of XEMC Windpower Co., Ltd. is used in this paper to analyze the yaw data of the No. 5 wind turbine of Baozhong Mountain Wind Farm in Hunan Province. In order to evaluate the power generation and the damage equivalent load (DEL) of the yaw bearing during the yaw process, an economic model predictive control (EMPC) yaw strategy based on light detection and ranging (LIDAR) was proposed. EMPC takes the yaw error and wind speed as the disturbance of the cost function (CF) at the same time by establishing the yaw bearing DEL look-up table in advance. Discrete adaptive yaw speed control set is proposed to adapt to different wind conditions sets. Finally, the effectiveness of EMPC is verified using the model of XE112-2000 wind turbine in simulation software Bladed.

The positioning systems’ high accuracy and reliability are crucial enablers for various future applications, including autonomous shipping worldwide. It is especially challenging for the Arctic region due to the lower number of visible satellites, severe ionospheric disturbances, scintillation effects, and higher delays than in the non-Arctic and non-Antarctic regions. In regions up North, conventional satellite positioning systems are generally proposed to be utilized, together with other situational awareness systems, to achieve the necessary level of accuracy. This paper provides a detailed review of the current state-of-the-art, satellite-based positioning systems’ availability and performance and reports high-level positioning requirements for the oncoming applications. In particular, the comparative study between three Global Navigation Satellite System (GNSS) constellations is executed to determine whether they are suitable for autonomous vessel navigation in the Arctics’ complex environment as the two most significant drivers for a reevaluation of the related satellite constellations. This work analyzes the ongoing research executed in different (inter-) national projects focused on Galileo, Global Positioning System (GPS), and GLObal NAvigation Satellite System (GLONASS). Based on the literature review and the simulation campaign, we conclude that all the convectional constellations achieve an accuracy of fewer than three meters in the analyzed Arctic scenarios. It is postulated that other complementary positioning methods should be utilized to improve accuracy beyond this limit. Finally, the study emphasizes existing challenges in the Arctic region regarding the localization and telecommunication capabilities and provides future research directions. publishedVersion

With the development of manufacturing industry, the high precision size measurement for the holes on the surface of mechanical parts is required. A method for measuring size of the hole based on double projection is proposed in this paper. The CCD camera separately collects grating images which projected onto the part surface in two directions. The phase-shift method is used to calculate the phase of grating images in two directions, and the reconstructed three-dimensional shapes in the two directions are merged to measure the hole dimension. Experiment results show that the measurement accuracy of the thickness is 0.04 mm, and the hole size can reach to 0.1 mm.

This paper presents a lane change decision algorithm for predictive decision-making for an autonomous vehicle using a Recurrent Neural Network (RNN) with a Bidirectional Long Short-Term Memory (Bi-LSTM) cell. The proposed decision-making algorithm was trained and validated by driving data collected by vision, laser scanners, and chassis sensors of autonomous vehicles. The input features for the Bi-LSTM based RNN consist of the clearance and relative velocity with the surrounding target vehicles, lane measurements, and the velocity of the autonomous vehicle. The output features are configured to generate the probability of three maneuvers, left lane change, right lane change, and lane-keeping. The Bi-LSTM based RNN is configured to decide in advance two seconds before lane changes by using two seconds of observation. The collected 20,108 datasets were accumulated in global coordinates. After processing and resampling the collected datasets, 1,120, 320, and 160 datasets were generated to train, validate, and test the Bi-LSTM based RNN. The proposed algorithm was evaluated by a case study and a driving data-based prediction accuracy analysis. The results of the predictive lane change decision by the proposed algorithm have been shown to be more accurate and similar to a driver than previous approaches.

The increase in the scale of distribution networks significantly reduces the efficiency of intelligent planning for distributed generation (DG). To improve the efficiency of intelligent DG planning and avoid the impact of uncertainty concerning renewable energy on it, this paper proposes a sensitivity index for the bus-embedded multi-objective estimation of distribution algorithm (MEDA) based on the semi-invariant probabilistic power flow approach to achieve an optimal solution. The sensitivity indices of the buses are comprehensively enabled to obtain a new index and determine their sensitivity sequences based on the area grey incidence decision-making method. Subsequently, according to the uncertainty of wind turbine generators and photovoltaics, a probability model is established, and the semi-invariant method is used to solve for the probabilistic power flow according to a correlation model. Finally, the sensitivity of the proposed bus-embedded MEDA to enhancing the efficiency of the solution is examined. The optimal DG allocation scheme is obtained with the goal of achieving the lowest total cost in the planning year. Finally, the feasibility and effectiveness of the proposed model and method are verified using simulations of the IEEE 33-bus, IEEE 69-bus, and IEEE 118-bus test systems.

Two-terminalt and tandem solar-cells have a high efficiency of power conversion. One of their main limitations is the operating current density as the two-terminal tandem solar-cell is equivalent to electrically connected series subcells. Increasing the top absorber layer’s thickness will lead to an increase in the top subcell current and a decrease in the bottom subcell’s current. The subcell with the minimum current forces the tandem cell to operate at its value, limiting the overall performance. In this paper, a proposed solution for such a problem is introduced using a bottom subcell consisting of germanium-telluride (GeTe), which gives a high current and matches the top subcell at a thicker absorber layer. A proposal of three different tandem cells with perovskite (MAPbI3)/CIGS, perovskite (MAPbI3)/GeTe, and perovskite (MAPbI3-xClx)/GeTe have been presented. The proposed perovskite (MAPbI3)/CIGS has an efficiency of 30.52%, whereas the replacement of the CIGS bottom subcell by GeTe led to a significant enhancement of the efficiency to reach 35.9%. High efficiency of 41.7% is obtained by replacing the perovskite (MAPbI3) top subcell with perovskite (MAPbI3-xClx). A modified numerical algorithm is proposed to obtain the optimum thickness of the top subcell to achieve higher power-conversion efficiency. The performance evaluation and simulation of the designed tandem cells were carried out using SCAPS-1D. The temperature effects on the proposed cells have been encountered in simulation. The results show that the proposed tandem solar-cells have comparable performance and higher efficiencies relative to the published works.

In the future, the population’s aging problem will become increasingly serious. For medical care, action-assisted vehicles are one of the solutions. A wheelchair is a common mobility assistance equipment, and the safety of the user is an important issue. For the ordinary wheelchairs, most of them use a fixed hand brake or disk brake to allow the users to stop the wheelchair. However, improper use of brakes tends to cause slippage, and there is no concept of the anti-lock braking system (ABS) for the wheelchair. Therefore, this paper applies this idea to intelligent wheelchairs with the aim of enhancing the safety of users in wheelchairs. The ABS architecture includes the core algorithms of the adaptive fuzzy-neural inference system and the friction coefficient estimation system. The friction coefficient estimation system uses a particle filter to quickly adapt to a non-linear state and unknown environment. The system provides more accurate braking control for the ABS according to the range of the friction coefficient based on the pavement type. In ABS, it uses the gyroscope to detect the acceleration and wheelchair angle information and then calculates the parameters of the wheelchair. The user clicks the stop command on the wheelchair to activate the brake system to achieve a simple and efficient mode of operation. It can efficiently reduce the braking time and braking distance and enhance the riding safety of wheelchairs.

To deal with the problem where each instance is associated with multiple labels, a lot of multi-label learning algorithms have been developed in recent years. Some approaches have been proposed to select label-specific features to utilize discriminate features for multi-label classification. Although label correlation has been considered in learning label-specific features, the critical correlation among instances was less taken into account. In this paper, we proposed a new approach called multi-label learning with label-specific features using correlation information (LSF-CI) to learn label-specific features for each label with the consideration of both correlation information in label space and correlation information in feature space. In the LSF-CI, the instance correlation in feature space is computed by a probabilistic neighborhood graph model, and label correlation in label space is computed by cosine similarity. For multi-label data, the LSF-CI has the capability to select Label-specific features for each label as well as classify an unseen instance into a set of relevant labels. To validate the effectiveness of LSF-CI, we conducted comprehensive experiments on eight multi-label datasets. The experimental results demonstrate that the LSF-CI is capable of selecting compact label-specific features, and achieving a competitive performance in comparison with the performances of the existing multi-label learning approaches.

This paper proposes a novel deep learning architecture for estimating the remaining useful lifetime (RUL) of industrial components, which solely relies on the recently developed transformer architectures. The RUL estimation resorts to analysing degradation patterns within multivariate time series signals. Hence, we propose a novel shared temporal attention block that allows detecting RUL patterns with the progress of time. Furthermore, we develop a split-feature attention block that enables attending to features from different sensor channels. The proposed shared temporal attention layer in the encoder fulfils the goal of attending to temporal degradation patterns in the individual sensor signals before creating a shared correlation across the feature range. We develop two transformer architectures that are specifically designed to operate with multivariate time series data based on these novel attention blocks. We apply the architectures to the well known C-MAPSS benchmark dataset and provide various hyperparameter studies to analyse their impact on the performance. In addition, we provide a thorough comparison with recently presented state-of-the-art approaches and show that the proposed transformer architectures outperform the existing methods by a considerable margin. CA extern

With the explosive growth of information and communication technology and its services, some popular Websites currently generate an enormous amount of Internet traffic. A content delivery network (CDN) would then become imperative for supporting such services efficiently. In this paper, we propose joint optimizing approaches for replica server placement, content caching in selected servers, and content request load assignment among the servers, aiming to minimize the ratio of unserved content request load when the network resources and server capacity are both limited. For this, we develop a mixed integer linear programming (MILP) optimization model. To mitigate the computational complexity of the MILP model, we further decompose the optimization problem into three sub-problems, including: 1) choosing the replica server nodes optimally; 2) deciding the content items to be cached in the replica servers; and 3) allocating the content request loads from users onto different servers. For these sub-problems, we develop corresponding heuristic algorithms and show that the proposed approach is not only efficient but also performs very close to the MILP model. We also find that a number of system limitations, such as different numbers of replica servers placed, link capacity, server processing capacity, and server storage capacity, jointly affect the performance of the CDN. Saturation trends are observed on the performance, which indicate that as long as sufficient resources have already been provided, augmenting the resources further may only lead to marginal additional performance improvement.

This paper presents a novel single-phase to single-phase multiconverter topology that can be applied in multiple areas. The proposed multiconverter is designed with only two soft power semiconductor switches (e.g. MOSFET or IGBT), four power diodes and a center-tapped transformer which makes it more compact in size, decrease the gate driving complexity, reduce the total equipment costs and enhance the energy conversion efficiency with minimized losses. Furthermore, the utilization of the transformer in the proposed converter mitigates the multiple AC source requirement problems and provides galvanic isolation which increases the reliability of the converter. Moreover, the presented multiconverter is applicable in various areas including electric traction as a speed controller, induction heating, AC and DC variable power supplies, etc. which signify the competence of this converter in energy conversion appliances. However, a comparative analysis of the offered converter with the existing AC-AC converters is also introduced in this paper with respect to the number of components, equipment costs, gate driving complexity, and application areas. In order to evaluate the performance of the proposed multiconverter, the simulation-based results carried out in MATLAB/Simulink are presented and analyzed in this paper with proper descriptions. Finally, a scaled-down prototype is developed in the laboratory to validate the simulation results and the feasibility of the proposed multiconverter.

In view of the instability and complexity of passenger flow change in urban rail transit, it is the key and the difficult point to use the prediction model to get more accurate number of short-term passenger flow. In view of this, this study proposes a hybrid forecasting model W-KELM, which combines wavelet transform (WT) and kernel extreme learning machine (KELM). The main idea of the model is to decompose passenger flow data into high-frequency and low-frequency sequences through WT and Mallat algorithm, and then use KELM approach to learn and forecast signals with different frequencies. Finally, different prediction sequences are reconstructed using WT. Through a case study of Beijing metro, we test the effectiveness of the model. The result shows that the W-KELM model has good prediction accuracy. In addition, this paper compare the prediction result of W-KELM model with those of BP neural network model, the single KELM method, and the hybrid model based on WT and BP neural network. It shows that the W-KELM model can effectively improve the prediction accuracy. Thus, providing a more accurate and real situation for monitoring and early warning of urban rail transit.

The use of Markov random fields (MRFs) is a common approach for performing image segmentation, where the problem is modeled using MRFs that incorporate priors on neighborhood nodes to allow for efficient Maximum a Posteriori inference. These local MRF models often result in smoothed segmentation boundaries, since they penalize the assignment of different labels to neighboring pixels and are limited in the use of long-range interactions. Although recent work on fully connected random fields and deep random fields has shown to be very promising in addressing these issues, both streams of approaches face certain limitations, which could affect inference performance and computational tractability. In this paper, we introduce the concept of deep randomly connected conditional random fields DRCRF, which fuse fully-connected random fields and deep random fields together to obtain benefits from long-range interactions while allowing for efficient inference using arbitrary potential functions. Leveraging random graph theory, the concept of stochastic cliques is incorporated into a deep CRF structure to take better advantage of long-range interactions while maintaining computational tractability. The experimental results demonstrate that the proposed DRCRF framework outperforms existing fully connected CRF frameworks and provides results comparable to the principled deep random field framework, which is among the state of the art in random field frameworks for image segmentation.

Mobile devices regularly broadcast WiFi probe requests in order to discover available proximal WiFi access points for connection. A probe request, sent automatically in the active scanning mode, consisting of the MAC address of the device expresses an advertisement of its presence. A real-time wireless sniffing system is able to sense WiFi packets and analyse wireless traffic. This provides an opportunity to obtain insights into the interaction between the humans carrying the mobile devices and the environment. Susceptibility to loss of the wireless data transmission is an important limitation on this idea, and this is complicated by the lack of a standard specification for real deployment of WiFi sniffers. In this paper, we present an experimental analysis of sniffing performance under different wireless environments using off-the-shelf products. Our objective is to identify the possible factors including channel settings and access point configurations that affect sniffing behaviours and performances, thereby enabling the design of a protocol for a WiFi sniffing system under the optimal monitoring strategy in a real deployment. Our preliminary results show that four main factors affect the sniffing performance: the number of access points and their corresponding operating channels, the signal strength of the access point and the number of devices in the vicinity. In terms of a real field deployment, we propose assignment of one sniffing device to each specific sub-region based on the local access point signal strength and coverage area and fixing the monitoring channel belongs to the local strongest access point.

In pantographs, current collector strips transmit the electrical energy they receive from the catenary to the locomotive and provide the necessary power for the locomotive's movement. In order for the current collector strips to transmit electricity to the locomotive in a healthy way, their surface must be smooth. Wear on the surface of the current collector strips reduces conductivity and can create arcs, endangering the health and safety of the pantograph and catenary system. In this paper, a Convolutional Neural Network (CNN) architecture is developed to detect wear on the current collector strips. Images obtained from pantographs used on railways were created with a clean and improved data set using Hough Transform and Power Law Transform. The created dataset contains 909 pantograph images. This dataset was trained and tested with both the developed CNN architecture and classic deep learning architectures (ResNet50, VGG16). The experimental results show that the developed CNN architecture's training results and test results are more successful than classical architectures.

This research paper proposes a multi-band frequency selective surface (FSS) director using the aperture interdigital technique. The FSS unit cell has been designed based on a basic bandpass filter (BPF) connected with an interdigital structure. With the proposed structure, the FSS unit cell size can be reduced from $\lambda $ /2 to $\lambda $ /8 caused by the slow-wave effect of interdigital capacitance loaded at the end of the transmission line in the unit cell structure. Moreover, the capacitance loaded at the end of the transmission line can control the second and the third resonance frequencies to resonate as required. The unit cell has been designed at the fundamental frequency of 1.8 GHz and the controlled second and third resonance frequencies of 3.7 GHz and 5.2 GHz, respectively. The operating frequencies of the proposed FSS unit cell cover LTE band (0-1.9 GHz), Wi-MAX band (3.22GHz-4.7GHz), and WLAN band (5.15 GHz-5.95 GHz). The size of the unit cell is very compact which is 11.53 mm $\times10.77$ mm. The designed unit cells are then connected as an array of $8\times 8$ , resulting in an overall size of 92.31 mm $\times86.16$ mm. This array has been used as a director for dipole antennas designed at the frequencies of 1.8 GHz, 2.45 GHz, 3.7 GHz, and 5.2 GHz. From the simulation, the dipole antennas with the proposed FSS director have directional gains of 4.52 dB at 1.8 GHz, 3.83 dB at 3.7 GHz, and 3.29 dB at 5.2 GHz, respectively. Measurement results show the directional gains of 3.67 dB at 1.8 GHz, 3.16 at 3.7 GHz, and 3.42 dB at 5.2 GHz, respectively. The proposed FSS director has properties of multi-band operation with a compact size that can be developed for antenna covers and radomes of multi-band wireless communications.

To ensure a long-time stable operation of the rolling bearing, it is important to accurately assess their working performance, especially the incipient degradation based on the massive service process data. As a new and effective tool, deep learning model is applied widely in the field of fault diagnosis but limited to rare labeled data. In this paper, a bearing performance assessment method based on signal component tracking is proposed to realize the bearing degradation detection. More general features are obtained by local convolution operation to represent the local characteristics in the spectrum or time–frequency distribution of vibration signal, which follows the forward features mapping process of the convolutional neural network (CNN). Then, a novel quantification criterion based on the comparison of those local features is used to provide the selection strategy of optimal fault components. The proposed method takes into account the abnormal information in degradation monitoring and utilizes it to achieve bearing incipient fault diagnosis. The experimental results prove that the features extracted by the proposed method possess high recognition efficiency when being used in incipient failure detection and diagnosis.

Understanding how careers evolve is very important for us, so we can make better professional decisions; for companies, so they can better plan their internal organization career progressions; and also for the governments, so they can better plan their labour market and economic policies. In this paper, we model a career progression as a graph, called MCar, in which nodes represent occupations and edges correspond to the flow of professionals among occupations. Based on MCar, we introduce some key concepts to investigate career evolution, namely, career boundary, occupational poles, and occupational islands. Finally, we propose the use of community detection techniques on a real database with millions of professional backgrounds to objectively identify career boundaries in Brazil and to study the topologies of the resultant graph. The results obtained provide a quantitative basis for career models, showing, for instance, the presence of hubs suggesting that less regular careers are common.

Computer-aided detection and diagnosis in ECG signals for heart diseases are gaining increasing attention. However, developing and selecting the highly performing diagnostic model suitable for clinical implications is still challenging. In this paper, we proposed a combined network of convolutional neural network (CNN) and Recurrent Neural Network (RNN), designed for the classification of ECG heart signals for diagnostic purpose. The proposed network consists of 2 convolutional layers with 5 × 5 kernels and ReLU activations, followed by 4 residual blocks, 2 bidirectional long short-term memory (biLSTM) layers, as well as 2 fully connected layers. Each residual block involved the structure of a Squeeze-and-Excitation Network (SENet) with lightweight property to recalibrate the feature map of the network. The last dense layer has 5 outputs, equivalent to the classes considered: non-ectopic beats, supraventricular ectopic beats, ventricular ectopic beats, fusion beats, and unknown beats. To train and evaluate the combined CNN and RNN, we transferred the knowledge acquired on beat classification tasks in 2017 PhysioNet/CinC Challenge to that in PhysionNet's MIT-BIH dataset. The developed network achieved a recognition sensitivity of 95.90%, accuracy = 95.90% and specificity = 96.34% with classification time of single sample = 6.23 s in detecting 5 ECG classes. A comparative analysis proved the high performance of the proposed combined CNN and RNN against previous methods, demonstrating the potential of our proposed network in the analysis of beat patterns. The proposed model can be applied in cloud computing or implemented in mobile devices to evaluate cardiac health with the highest precision.

The use of supplementary controllers for mitigating subsynchronous interaction (SSI) in series compensated DFIG-based wind farms is quite promising due to high effectiveness and low cost. Implementation of such a controller requires effective communications between individual turbines and the wind farm controller, where the control performance is very much affected by the communication delays involved. This paper delivers the first detailed analysis on the impact of communication delays on SSI damping controller performance. A novel algorithm is proposed to calculate the stability delay margin (SDM) of the closed-loop system based on Rekasius Substitution and Guardian Map Theorem with the advantage of reduced computational burden particularly for high-order systems. Based on the proposed algorithm, the impacts of wind farm operating conditions and turbine control parameters on the SDM are investigated. To strengthen the SSI damping controller performance against communication delays, a Smith predictor scheme is also developed. The effectiveness of the proposed delay analysis framework and Smith predictor scheme based delay immune controller is validated through Electromagnetic Transient (EMT) simulations on a realistic test system with multiple series capacitor compensated lines considering various operation conditions.

Internet of Things (IoT) has been the driving force for many smart city applications. The huge volume of IoT data generated from these applications require efficient processing to get the insight, which poses significant difficulty. Data mining and machine learning (DM) algorithms are used to minimize such difficulty. However, it is still very challenging to select a particular DM algorithm that can process a dynamic IoT dataset based on some application-specific goals to achieve better accuracy. This paper proposes a knowledge-driven framework that considers the knowledge of datasets, available DM algorithms, and application goals to select the suitable DM algorithm for performing a target data processing task. This work considers data from cultural domain, health domain, and transportation domain in the experiment. The results show that the proposed approach dynamically selects the best-suited DM algorithms for the available datasets and target goals that exhibits satisfactory performance in obtaining accurate results compared to the existing work. The proposed approach not only provides flexibility in conducting dynamic IoT data mining tasks, but also reduces the complexity that would otherwise be necessary while adopting the traditional data mining approaches.

Coded caching is widely regarded as an important tool to reduce pressure on the data transmission during the peak traffic times in heterogeneous wireless networks. In this paper, we will provide two concatenating constructions of coded caching schemes based on the schemes derived by Shangguan et al. in 2018 and Yan et al. in 2017, respectively. Then we show that our new schemes have good performance in reducing packet number. Furthermore, in some case, we present that our schemes have better performances on user number, memory ratio, packet number and transmission rate than the schemes obtained by Chittoor et al. in 2019 and Krishnan in 2018.

Non-orthogonal multiple access (NOMA) has been recognized as an essential technology for beyond fifth-generation (beyond-5G) wireless networks to increase connectivity, spectral efficiency, cell-edge throughput, and user fairness. In this paper, we propose two subcarrier-user assignment algorithms (SUAAs) for the downlink NOMA system to enhance the spectral efficiency, the fairness, the data rate of weak users, and the outage probability. The assignment order of the first proposed SUAA is based on the worst subcarrier first (WSF) to avoid selecting a user with the worst channel gain with any subcarrier and called (WSF-SUAA). On the other hand, the second proposed SUAA is based on spectral efficiency maximization (SEM) and called (SEM-SUAA), but requires exhaustive search. The assignment process of both algorithms is based on making the channel gain of the selected paired users per subcarrier as high as possible to increase the data rate of each user. Besides, the assignment process of strong users for all subcarriers is performed before the assignment process of weak users to increase the total system sum-rate. It is exposed throughout the simulation that the two proposed SUAAs can attain significant improvement in the total spectral efficiency, weak user data rate, outage probability, and user fairness compared to the existing algorithms. While the performances of the two proposed SUAAs are convergent, the computational complexity of WSF-SUAA is significantly lower than that of SEM-SUAA and slightly higher than that of the existing algorithms.

Covariance analysis describing function technique is a conventional method to solve the performance analysis of the nonlinear missile guidance system. Aiming at the faultiness of covariance analysis describing function technique and its improved method, this paper proposes a new method to solve the mean and covariance transfer issue by using the high-degree spherical-radial cubature rule. In terms of simulation models, the nonlinear missile guidance system contains the saturation link to limit the lateral acceleration command and the initial heading error is introduced. In addition, the zero-mean target evasive maneuver with exponentially-decaying autocorrelation function is chosen as the input of the missile guidance system. Furthermore, compared with the existing methods, the proposed method has higher computational efficiency, accuracy and better applicability in high-dimensional nonlinear systems by theoretical derivation and simulation results from the mean and covariance values of relative lateral separation. Finally, the presented analysis method can be used as a general method for the design and statistical analysis of multidimensional nonlinear systems.

Nonlinear Model-based Predictive Control (NMPC) is a relevant research area having applications in the industrial sector. Traditionally, in this technique, gradient descent algorithms have been used to solve the related optimization problem. More recently, bio-inspired meta-heuristics have also been applied to this problem. However, only a few works have been devoted to testing solvers that use parameter control with self-adaptive traits, which allows mitigating the problem of offline parameter tuning in bio-inspired approaches. In this paper, we propose the novel Adaptive Modified Grey Wolf Optimization (AMGWO) and the Adaptive Moth-Flame Optimization (AMFO), for solving Nonlinear Model-based Predictive Control (NMPC) problems. To achieve this, a mechanism for individual leaders weighting and a crossover operator are introduced in AMGWO, and a simple self-adaptive parameter technique is applied in both meta-heuristics. The improved solvers are tested to perform the swing-up of a single inverted pendulum and attitude control of a satellite, which are nonlinear problems relevant for assessing control performance. Nonparametric statistical tests are applied to compare the improved meta-heuristics optimization outcomes with other five meta-heuristics, which shows that the self-adaptive parameter technique can significantly improve the performance when applied as an NMPC solver, as the AMFO and AMGWO statistically outperform or performs as well as all algorithms compared in both the pendulum and satellite control, respectively. This is important as improving the optimizer efficiency will lead to more accurate control and enable rapid hardware implementation.

Human behaviors are known to spread through social contact. The diffusion process on social networks has also been leveraged to understand the spread of undesirable contagion. The contagion of malicious or even criminal behaviors in online social networks is just beginning to attract attention. Here, we study the social contagion problem of cheating behavior found in the massively multiplayer online role-playing game (MMORPG) that provides a lifelike environment with rich and realistic user interactions. Because cheating users boast an abnormal thus conspicuous degree of success, it has a strong chance of being noticed by their friends and leading them to cheat themselves. To detect and prevent cheating, it is beneficial to understand this dynamic as a contagion problem. In this paper, we show the existence of the contagion of cheating. We then explore various possible social reinforcement mechanisms after introducing several factors to quantify the effect of social reinforcement on the contagion and analyze the dynamics of bot diffusion in an extensive user interaction log from a major MMORPG.

In this paper, we describe improvements to the particle swarm optimizer (PSO) made by the inclusion of an unscented Kalman filter to guide particle motion. We show how this method increases the speed of convergence, and reduces the likelihood of premature convergence, increasing the overall accuracy of optimization. We demonstrate the effectiveness of the unscented Kalman filter PSO by comparing it with the original PSO algorithm and its variants designed to improve the performance. The PSOs were tested firstly on a number of common synthetic benchmarking functions and secondly applied to a practical three-dimensional image registration problem. The proposed methods displayed better performances for 4 out of 8 benchmark functions and reduced the target registration errors by at least 2mm when registering down-sampled benchmark brain images. They also demonstrated an ability to align images featuring motion-related artifacts which all other methods failed to register. These new PSO methods provide a novel, efficient mechanism to integrate prior knowledge into each iteration of the optimization process, which can enhance the accuracy and speed of convergence in the application of medical image registration.

Intelligent Transportation Systems (ITS) research and applications benefit from accurate short-term traffic state forecasting. To improve the forecasting accuracy, this paper proposes a deep learning based multitask learning Gated Recurrent Units (MTL-GRU) with residual mappings. To enhance the performance of the MTL-GRU, feature engineering is introduced to select the most informative features for the forecasting. Then, based on real-world datasets, numerical results show that the MTL-GRU can well estimate traffic flow and speed simultaneously, and performs better than other counterparts. Experiments also show that the deep learning based MTL-GRU model can overpower the bottleneck caused by enlarging training datasets and continue to gain benefits. The results suggest the proposed MTL-GRU model with residual mappings is promising to forecast short-term traffic state.

In this paper, an efficient digital predistortion (DPD) architecture utilizing over-the-air (OTA) feedback is presented to linearize phased array transmitters. It places one additional observation antenna (OA) in the far field to capture the combined signal of the array outputs and uses the combined signal to build a nonlinear model that describes the joint nonlinear distortions of all the power amplifiers (PAs) in the phased array. Then, a corresponding DPD model is extracted by the typical indirect learning method to linearize the phased array. After introducing this DPD architecture, the relationship between the OA position and the performance of the extracted DPD model is specifically explored, and two significant results are found. First, when the OA is perfectly colocated in the main beam direction, the DPD coefficients extracted with an arbitrary steering angle for the phased array are also applicable to any other steering angle. Second, when the OA is not perfectly placed in the main beam direction, the linearization performance for the extracted DPD coefficients will degrade as the OA placement mismatch increases, and the phased array has an optimum steering angle at which the OA placement mismatch has a minimal influence on the DPD performance. Based on the above two results, a new DPD training strategy is proposed, which extracts only one set of DPD coefficients by making the phased array point at the optimum steering angle. The advantages of the proposed strategy are that the OA can be placed in a relatively wide range of positions and that the extracted coefficients will be applicable to any other steering angle. The two results and the proposed DPD training strategy are verified by simulations, which are conducted on a $1\times 4$ uniform linear array (ULA) and a $4\times 4$ uniform rectangular array (URA).

Frame arrival detection is the first crucial task for a digital communication receiver. In this paper, we propose a new frame arrival detection method based on the concept of matched filtering. The proposed training sequence consists of three repeated subsequences, each including a few repeated segments and exhibiting high sparsity in frequency domain. The proposed detection method includes the following two stages. The first matched filtering stage employs the subsequence as the filter coefficients, which matches the frequency sparsity of the received subsequence and could greatly improve the output signal to noise ratio. The second stage adopts delayed autocorrelation on the filtered signal to detect the presence of training sequence. It is demonstrated that the proposed method could outperform the conventional methods in terms of both missed detection and false alarm probability. We derive the theoretical analysis for both missed detection and false alarm performance. We further extend our performance analysis and numerical evaluation to the system with carrier frequency offset (CFO). The results illustrate accuracy of the analytical results and robustness of the proposed method against practical levels of CFO.

N-1 static security analysis (SSA) is an important method for power system stability analysis that requires solving N alternating-current power flows (ACPF) for a system with N elements to obtain strictly accurate results. Past researches have shown the potential of accelerating these calculations using an iterative solver with graphics processing unit (GPU). This paper proposes a GPU-based N-1 SSA algorithm with the preconditioned conjugate gradient (PCG) method. First, a shared preconditioner is selected to accelerate preprocessing of the iterative method for fast decoupled power flow (FDPF) in N-1 SSA. Second, it proposes a GPU-based batch-PCG solver, which packages a massive number of PCG subtasks into a large-scale problem to achieve a higher degree of parallelism and better coalesced memory accesses. Finally, the paper presents a novel GPU-accelerated batch-PCG solution for N-1 SSA. Case studies on a practical 10828-bus system show that the GPU-based N-1 SSA algorithm with the batch-PCG solver is 4.90 times faster than a sequential algorithm on an 8-core CPU. This demonstrates the potential of the GPU-based high-performance SSA solution with the PCG method under a batch framework.

In this paper, the design of reduced-order and full-order interval observers for discrete-time system is investigated. First, the original system is transformed into a descriptor system with the ability to estimate the unknown noise of sensors. Next, a reduced-order interval observer design method for descriptor system is presented. The state matrix of the system is obtained by linear transformation, and a Sylvester equation is constructed to find the appropriate observer gain matrix. Then, an existing design method of full-order interval observer is introduced in this paper. Finally, a simulation example is given to demonstrate that the reduced-order interval observer has more accurate interval estimation results than the full-order interval observer.

Nowadays, Wireless Sensor Networks (WSNs) are playing a vital and sustainable role in many verticals touching different aspects of our lives including civil, public, and military applications. WSNs majorly consist of a few to several sensor nodes, that are connected to each other via wireless communication links and require real-time or delayed data transfer. In this paper, we propose an autonomous Unmanned Aerial Vehicle (UAV)-enabled data gathering mechanism for delay-tolerant WSN applications. The objective is to employ a self-trained UAV as a flying mobile unit collecting data from ground sensor nodes spatially distributed in a given geographical area during a predefined period of time. In this approach, two Reinforcement Learning (RL) approaches, specifically Deep Deterministic Gradient Decent (DDPG) and Q-learning (QL) algorithms, are jointly employed to train the UAV to understand the environment and provide effective scheduling to accomplish its data collection mission. The DDPG is used to autonomously decide the best trajectory to adopt in an obstacle-constrained environment, while the QL is developed to determine the order of nodes to visit such that the data collection time is minimized. The schedule is obtained while considering the limited battery capacity of the flying unit, its need to return the charging station, the time windows of data acquisition, and the priority of certain sensor nodes. Customized reward functions are designed for each RL model and, through numerical simulations, we investigate their training performances. We also analyze the behavior of the autonomous UAV for different selected scenarios and corroborate the ability of the proposed approach in performing effective data collection. A comparison with the deterministic optimal solution is provided to validate the performance of the learning-based approach.

Deep learning has achieved great success in the field of computer vision, and the precision in image classification and image detection has surpassed humans. Therefore, this paper combines deep learning and medical image segmentation, focusing on how to improve the accuracy and speed of segmentation algorithm of medical exercise rehabilitation image. Aiming at the shortcomings of traditional medical image recognition methods, a medical exercise rehabilitation image segmentation algorithm based on hierarchical features of convolutional neural networks is proposed, this paper calls it as hierarchical features of convolutional neural networks (HFCNN). The algorithm firstly samples the convolution output of multiple layers in the convolutional neural network to a unified scale and combines them to construct a hierarchical feature. This hierarchical feature combines the structural information of objects contained in the shallow layer of the network with the semantic information of objects contained in the deep layers of the network, so it has a strong ability to express. Secondly, the image can be segmented into multiple super pixels by the super pixel segmentation algorithm. The classifier is trained using the hierarchical features of the super pixel, and then the classification result is mapped back to the pixel. Finally, a fully connected conditional random field algorithm including one-potential potential energy and paired potential energy is constructed. The corresponding energy function is used to smooth the classification result of the pixel, and the regional consistency and continuity of the pixel mark are improved. Compared with many classical convolutional neural network algorithms, this algorithm not only accelerates the network convergence speed, shortens the training time, but also significantly improves the accuracy of segmentation algorithm of medical exercise rehabilitation image, showing good practical value.

This article analyzes the serum electrolyte disturbances of patients through deep learning algorithms. Among the 104 patients with electrolyte disturbances, 6 cases of serum potassium, sodium, chloride, calcium, phosphorus, and magnesium electrolyte disturbances have occurred, the proportion of occurrence The order is sodium > chlorine > calcium > potassium > phosphorus > magnesium. This paper proposes a deep learning algorithm for serum electrolyte disorder, and analyzes and implements the functions at various levels according to the characteristics of the Hadoop framework. The system includes electronic medical records shared storage, distributed realization of definite learning algorithms, classification and recognition of myocardial ischemia by deep learning, and Web system assisting doctor diagnosis, which lays the foundation for the construction of serum electrolyte disorder scientific research data center. In order to explore this relationship, without artificially extracting features, a deep learning model of convolution and long-term and short-term memory circulation neural network cascade was proposed to determine the positive or negative myocardial ischemia by classifying serum disorders. Conduct clinical experiments, including patients with suspected coronary heart disease and coronary angiography as the research object, taking coronary angiography results as the detection standard. Experimental results show that the model has an accuracy of 89.0% for detecting myocardial ischemia, a sensitivity of 91.7%, and a specificity of 81.5%. A linear combination model of CNN and LSTM is proposed to classify and recognize serum electrolyte disorders. Determine the learning theory to dynamically model the ST-T segment in the ECG to obtain serum electrolyte disturbance, which more vividly shows the changes in electrical information under myocardial ischemia. In order to reveal the relationship between serum electrolyte disturbance and myocardial ischemia, this paper builds a neural network model to learn and train serum electrolyte disturbance data to realize the classification of positive or negative myocardial ischemia. Tests on serum electrolyte disturbance data collected in clinical experiments show that this model can better achieve early detection of myocardial ischemia.

Directional modulation (DM) as a physical layer security technique has been studied from many different aspects recently. Normally all existing designs based on antenna arrays lead to varying weight coefficients for different antennas and for different signal symbols, which poses a particular challenge for feed circuits design in analogue implementation. In this paper, to reduce the implementation complexity, a constant magnitude constraint is proposed for the first time, and the resultant non-convex constraint can be modified to a convex form so that the problem can be solved conveniently by existing convex optimisation toolboxes. Design examples are provided to show the effectiveness of the proposed design.

Visual question answering is a challenging multimodal task, which has received increasing attention in recent years. One key solution to visual question answering is how to fuse the visual and textual features extracted from the image and questions, and thus, we can comprehensively employ the information from both modals and deliver correct answers. Bilinear pooling has been a powerful fusion approach owing to its exhausting interaction of each element of two modals, but its overuse of parameters limits its practical application. In this paper, we aim to retain the advantages of bilinear pooling for feature interaction and propose a novel multimodal feature fusion approach named multimodal local perception bilinear (MLPB) pooling, which can retain the second-order interactions between visual and textual features with limited learning parameters. To be specific, the MLPB utilizes local perception mechanism, which transforms the bilinear pooling between two high-dimensional raw features into multiple low-dimensional part features. To further reduce the computational cost, we propose to share the learning parameters of each local bilinear pooling. In this way, MLPB can achieve the complex interactions of the bilinear pooling without taking up too much computational resource. Extensive experiments show that the proposed method can achieve competitive or better performance than the state of the art.

This paper investigates the adaptive fault-tolerant control problem for a class of nonlinear systems with unknown control directions and uncertain parameters. To solve this control problem, the homogeneous domination method based on backstepping and fault-tolerant control scheme is applied to the uncertain nonlinear systems. First, a homogeneous domination-based adaptive control scheme using Nussbaum-type function is designed. Then, a fault-tolerant controller is proposed to deal with the actuator faults. The adaptive controller ensures that the closed-loop system is bounded, and all the states of the system asymptotically converge to 0. Finally, a simulation example is applied to demonstrate the effectiveness of the new designed adaptive fault-tolerant control scheme.

Robot-assisted cooperative rehabilitation training has shown superiority in helping the individuals with motion impairment problems to regain their motor functions. This paper presents the development and evaluation of a new cooperative training control scheme for an end-effector-type rehabilitation robot to provide upper extremity rehabilitation training with desired compliance and intensity. Firstly, the overall mechanical structure and real-time control system of rehabilitation robot are introduced. Secondly, an integral fuzzy sliding mode impedance control strategy combined with time-delay estimation (IFSMIC-TDE) is proposed to induce the active participation of patients and suppress impedance error. The IFSMIC-TDE approach is free from nonlinear robot dynamics, and it is designed to be robust to the external uncertainties and inherent chattering characteristics. After that, the closed-loop system stability with IFSMIC-TDE is proved based on the Lyapunov stability theory. Finally, experimental investigations are conducted to illustrate the effectiveness of the proposed rehabilitation robot and control algorithm. The comparison results indicate that the proposed IFSMIC-TDE can achieve better control performance and less impedance error. Besides, the interaction compliance and training intensity can be qualitatively adjusted via appropriate impedance parameters.

Non-orthogonal multiple access (NOMA) is a promising spectrally-efficient technology to satisfy the massive data requirement of the next-generation wireless communication networks. In this paper, we consider cooperative non-orthogonal multiple access (CNOMA) networks with relaying user selection, where a two-stage user selection scheme is adopted to optimize the received signal-to-noise ratio (SNR) of the weak user. We derive analytical expressions of the outage probability of the CNOMA users under two different types of channel estimation errors. Afterwards, we investigate the asymptotic characteristics the outage probability of the CNOMA systems in the high SNR regimes. Our results show that for the case of constant variance of the channel estimation error, the outage probability of each user is saturated and cannot be improved by increasing the transmit power of the access point. In contrast, there is no performance bottleneck of the outage probability for all the users when the variance of the channel estimation errors decease with SNR, whilst fully diversity gain is achieved by the weak user.

Although hyperspectral remote sensing images have rich spectral features, for small samples of remote sensing images, feature selection, feature mining, and feature integration are very important. A single model is difficult to apply to multiple tasks such as feature selection, feature mining, and feature integration during training, resulting in poor classification results for small sample classification of hyperspectral images. To improve the classification of small samples, a sequential joint deep learning algorithm is proposed in this paper. (In this algorithm, the deep features of multiscale convolution under an attention mechanism are integrated by using Bidirectional Long Short-Term Memory(Bi-LSTM) and AML.) First, we used principal component analysis (PCA) to reduce the dimensionality of the hyperspectral data and retain their key features. Second, the model uses an integrated attention mechanism to distribute the probability weight of the key input feature. Third, the model uses multiscale convolution to mine features after the distribution weight to obtain deep features. Fourth, the model uses bidirectional long short-term memory (Bi-LSTM) to integrate the convolution results at different scales. Finally, the softmax classifier is used to complete the classification of multiclass hyperspectral remote sensing images. Experiments were carried out on three public hyperspectral data sets, and the results proved that our proposed AML algorithm is effective, thus demonstrating powerful performance in the prediction of hyperspectral images (HSIs) of small samples.

In this paper, the capacitated location-routing problem with fuzzy demands (CLRP-FD) is considered, which simultaneously solves two problems: locating the facilities and designing the vehicle routes among the established facilities and customers. In the CLRP-FD, the capacities of the employed vehicles and established facilities cannot exceed, and the demands of the customers are assumed to be triangular fuzzy variables. To model the CLRP-FD, a fuzzy chance constrained programming approach is designed using fuzzy credibility theory. To solve this problem, a hybrid particle swarm optimization (HPSO) algorithm, which includes a stochastic simulation and a local search strategy based on the variable neighborhood search algorithm, is proposed. Finally, the influence of the value of the dispatcher preference index ( DPI ) on the total distribution cost is analyzed by conducting numerical experiments. To evaluate the efficiency of the proposed HPSO and the performance of the CLRP-FD model, the results obtained using the HPSO were compared with their corresponding lower bound provided using the CPLEX solver. Moreover, we also evaluated the performance of HPSO through computational experiments on some well-known benchmark CLRP instances. The numerical results show that the proposed HPSO is competitive, and can give a satisfactory solution in a reasonable amount of time.

Next basket recommendation with consideration of user sequential shopping behaviors plays a significant role in E-commerce to improve the user experience and service quality. Recently, recurrent neural networks (RNNs), especially attention-based RNN, have been widely adopted in the next basket recommendation. However, existing fixed attention mechanisms are not designed to model the dynamic and diverse characteristics of user appetites. In this paper, we propose an inter-basket and intra-basket adaptive attention network (IIAAN) for the next basket recommendation. Specifically, the inter-basket adaptive attention acts on all historical user baskets to model user's diverse long-term preferences, while the intra-basket adaptive attention is designed to act on item-level in the most recent basket to model user's dynamic and different short-term preferences. Then, we further integrate inter-basket and intra-basket adaptive attentions together to improve recommendation effectiveness. Finally, we evaluate the proposed model IIAAN using three real-world datasets from various E-commerce platforms. Our experimental results show that our model IIAAN significantly outperforms the state-of-the-art approaches for the next basket recommendation.

Currently, high dynamic range (HDR) videos with high resolution (HR) have become popular due to the display and the rendered technological advancements. However, making ultra-high definition (UHD) with HDR videos is expensive. The legacy low-resolution (LR) standard dynamic range (SDR) format is still largely used in practice. It is necessary to search for a solution to transform LR SDR videos into UHD HDR format. In this paper, we consider joint super resolution and learning inverse tone mapping an issue of high-frequency reconstruction and local contrast enhancement, and we propose an architecture based on a generative adversarial network to apply joint SR-ITM learning. Specifically, we include the residual ResNeXt block (RRXB) as a basic module to better capture high-frequency textures and adopt YUV interpolation to achieve local contrast enhancement. By adopting a generative adversarial network as a pivotal training mechanism, our designs show advantages in both integration and performance. Our code is now available on GitHub: SR-ITM-GAN.

In traditional wireless sensor networks (WSN), data is transmitted to base stations through multiple-hop. However, multi-hop transmission will cause problems such as energy holes, uneven energy consumption, and unreliable data transmission. Aiming at the defects of traditional networks, this paper proposes a scheme for collaborative data collection using multiple mobile nodes (MN) as sink nodes. In this scenario, an important issue is how to reasonably plan the data collection path of each mobile node under a series of constraints such as energy. This paper studies routing strategy and path planning. Firstly, a dynamic clustering algorithm is used to cluster the randomly arranged sensor nodes, and then a sensor node with higher energy is manually arranged at the virtual cluster center generated by the clustering algorithm as a cluster head node to establish a data collection cluster. Then the monitoring area is divided into several parts according to the number of MN, so that the MN traverses the cluster head nodes in the respective monitoring areas for data collection. At the same time, in order to enable the MN to complete data collection within the energy limit and improve the path balance, a path-based path equalization algorithm (PEABR) is proposed to adjust the path of the MN, further reduce and equalize the path length of the MN to satisfy the constraint conditions, and optimize the path planning scheme. Finally, simulation was carried out by using Matlab simulator, and the simulation experiments which were performed in a laboratory environment verified the feasibility and effectiveness of the algorithm.

With the popularity of Android smartphones, malicious applications targeted Android platform have explosively increased. Proposing effective Android malware detection method for preventing the spread of malware has become an emerging issue. Various features extracted through static and dynamic analysis in conjunction with machine learning algorithm have been the mainstream in large-scale malware identification. In general, static analysis becomes invalid in detecting applications which adopt sophisticated obfuscation techniques like encryption or dynamic code loading. However, dynamic analysis is suitable to deal with these evasion techniques. In this paper, we propose an effective dynamic analysis framework, called EnDroid, in the aim of implementing highly precise malware detection based on multiple types of dynamic behavior features. These features cover system-level behavior trace and common application-level malicious behaviors like personal information stealing, premium service subscription, and malicious service communication. In addition, EnDroid adopts feature selection algorithm to remove noisy or irrelevant features and extracts critical behavior features. Extracting behavior features through runtime monitor, EnDroid is able to distinguish malicious from benign applications with ensemble learning algorithm. Through experiments, we prove the effectiveness of EnDroid on two datasets. Furthermore, we find Stacking achieves the best classification performance and is promising in Android malware detection.

Emerging deep learning (DL) techniques have greatly improved pedestrian reidentification (PRI) performance. However, the existing DL-based PRI methods cannot learn robust feature representations owing to the single view of query images and the limited number of extractable features. Inspired by generative adversarial networks (GANs), this paper proposes a novel PRI method based on a pedestrian multiview GAN (PmGAN) and a classification recognition network (CRN). The PmGAN consists of three generators and one multiclass discriminator. The three generators produce pedestrian images from the front, side and back, while the multiclass discriminator determines whether the input image is a real image or a generated image. In addition to expanding the existing pedestrian datasets, the PmGAN can generate pedestrian images from front, side and back views based on a given query image and thereby increase the feature semantic space of the query image. To verify the performance of our method, the PmGAN was compared with mainstream pedestrian image generation models, and then the proposed method was contrasted with mainstream PRI methods. The results show that the proposed PmGAN greatly improved the performance of mainstream PRI methods. For example, the combination of the PmGAN and Pyramidal increased the mean average precision (mAP) on three common datasets by 1.2% on average. The research findings provide new insights into the application of multiview generation in PRI tasks.

Accurate remaining useful life (RUL) prediction under the noisy environment is a big challenge for the health management of modern industrial systems since the extraction of the accurate data structure from heavily corrupted data is difficult. In recent years, the kernel adaptive filter (KAF) has been widely adopted to solve the robust regression problem due to its low-complexity and high-approximation capability and robustness while the applications in battery RUL prediction are still few and far between. Thus, this paper is concerned with long-term RUL prediction using the KAF method. At first, concretely speaking, a robust KAF algorithm is derived based on the double-Gaussian-mixture (DGM) cost function, which is used to learn the capacity degradation mechanism from contaminated capacity data and so as to build the long-term prediction model. Second, a robust unscented Kalman filter (UKF) algorithm employing the DGM-based cost function is developed, which is then combined with the KAF-based prediction model to realize a more accurate and reliable prediction. Under the hybrid prognostic framework, the proposed UKF algorithm is applied to filter the noisy observations. When the observation data are inaccessible, the predicted data from the off-line trained KAF-based prediction model are adopted as the approximated value of the real observations for the UKF algorithm to optimize the prediction results and to provide the uncertainty representation. The experimental results reveal that the proposed method has great robustness when the measurements contain noise and large outliers, which makes it possible to get satisfactory prediction performance without preprocessing the data manually.

Hyperspectral transmission imaging may provide a means for rapid screening of breast tumors, but tissue has a strong nature of scattering, thus causing a great difficulty in identifying heterogeneity. In this paper, a combination of frame accumulation and deep learning was proposed to detect heterogeneity, and we designed the simulation experiment of collecting phantom images. On the basis of frame accumulation preprocessing, the heterogeneous detection is performed on multispectral images by using faster regions with convolutional neural networks (R-CNN) features and a single shot multibox detector (SSD), two typical detection frameworks of deep learning. The results show that the mean average precision (mAP) of faster R-CNN and SSD reach 90.8% and 95.1%, respectively, when three classes (including background) are detected with the help of the dataset provided in this paper, and the mAP of two frameworks both reach 99.9% when two classes (including background) are detected. The detection efficiency of the SSD is higher than faster, SSD’s detection speed can reach 50 fps, and the detection accuracy of the images after frame accumulation preprocessing is higher than that without frame accumulation processing. In summary, we validate the possibility of employing faster R-CNN and SSD to detect heterogeneity in multispectral images based on frame accumulation that improves image grayscale resolution, and it has a certain degree of reference significance for the application of deep learning in multispectral image detection.