Abstract:In order to tackle the dynamic scheduling problem of tow trains in mixed-model assembly lines, a scheduling approach is proposed based on the knowledge base and neural network. Firstly, the dynamic scheduling problem of material delivery in the automotive assembly line is formally described. The throughput of the assembly line and the total delivery distances are selected as components of the objective function. After that，the sample data of mixed-model assembly lines are generated by the Plant Simulation software and are used to train the neural network model offline. Finally, the trained neural network model and the knowledge base are adopted in the real-time scheduling process to select the optimal scheduling rule for tow trains. The experimental results indicate that the scheduling rules selected by the selection method proposed in the paper are mostly the optimal ones. The lower computational complexity of scheduling rules ensures the real-time performance of scheduling. It can cope well with changes in the dynamic environment, thus effectively improving the dynamic scheduling of tow trains.

Abstract:This paper designs an experimental scheme based on the driving simulation platform and collects driver's operation information and vehicle status information synchronously. Six characteristic parameters are selected to recognize the driving style. The principal component analysis (PCA) algorithm is used to extract the multi-feature parameters and the first three principal components are used as the input features of the driving style recognition model. The K-means method is used for data labeling. Based on the principles of supervised support vector machine (SVM) method and inductive multi-label classification with unlabeled data (iMLCU) approach, the driving style recognition models of SVM and iMLCU are established, respectively. By adjusting the trained dataset ratios between the labeled and the unlabeled data, the accuracy of driving style recognition between the two models is compared. The results show that iMLCU has better driving style recognition than SVM. The semi-supervised iMLCU model can improve the recognition ability of driving style by using unlabeled samples.

Abstract:Using DrivAer vehicle model, the mechanism of drag reduction and noise reduction of bionic non-smooth rear-view mirror cover is studied. Wind tunnel experiments verify the validity of LES (large eddy simulation) and k - ε simulation models，and show that the rear-view mirror can increase air resistance and noise. The bionic non-smooth structure is applied to the surface of rear-view mirror cover of the DrivAer vehicle model. The simulation results show that the reasonable application of bionic non-smooth structure can reduce the resistance of the vehicle by 5.9% and the loudness outside the side window by 19.4%; The bionic non-smooth structure promotes the formation of the vortex pad effect by changing the flow state of boundary layer, which reduces the energy loss of incoming flow, improves the stability of flow field, and then has a positive impact on the aerodynamic drag and noise of the vehicle.

Abstract:Piston wind is the main aerodynamic characteristics of the flow field in the evacuated tube，and thus grasping the basic characteristics and variation rules of the piston wind is the basis of a reasonable and effective control of the flow field of the tube. In this paper, the influences of piston air generation mechanism，aerodynamic characteristics and degrees of vacuum，blockage ratios, driving speeds and other conditions on the piston wind were analyzed and discussed by using the methods of computational fluid dynamics and dynamic meshing. This study finds that air in the pipeline is compressed and inflated during the running of the vehicle, producing compression and expansion waves and affecting the vehicle's driving resistance；Through a comprehensive analysis on a series of combined calculations for different degrees of vacuum，blockage ratios and driving speeds，it is found that the vehicle's driving resistance increases with the growth of the blocking ratio，pressure, and speed. Further，when the speed increases to a threshold, the increase of the vehicle’s driving resistance begins to slow down.

Abstract:A small volume and high stiffness periodic metal isolator composed of elastic plates and supporting columns is designed. Firstly, the static load test is carried out by using microcomputer controlled electronic testing machine, and the load-displacement characteristic curve is obtained. Based on the finite element method, a finite element model of periodic structure is established, and the static analysis is carried out under the same working conditions as the test. On this basis, the influence of the main dimension parameters on the stiffness is studied. The results show that the thickness of elastic sheet has great influence on the stiffness, while the inner diameter and the number of layers have little influence on the stiffness. Finally, by using the equivalent spring-mass model of the periodic structure isolator with multiple degrees of freedom, its dynamic equation is established by means of modal superposition method, and the expression of force transfer rate is deduced. A simulation is carried out based on the finite element analysis which is verified by experiment. The results show that the theory and simulation match well. When the load mass exceeds a certain value, the periodic structure can be simplified to single degree of freedom.

Abstract:Considering the two characteristics of activity splitting and resource window in the process of aircraft assembly，the model and algorithm of Resource Investment Problem on aircraft mobile production line were studied. Aiming at the situation that some activities have known splitting mode and splitting punishment，an improved genetic algorithm for solving this problem was designed. The traditional real value crossover operation was optimized，and a crossover method based on chromosome fitness value was proposed. Sensitivity analysis was carried out on the range of values of the relevant parameters. A mutation mechanism based on the probability of selection of activity start time was also proposed. For a scheduling scheme that satisfies the optimization conditions，combined with the position of the resource window，after judging whether the splitting activities can be re-scheduled and executed by selecting a new splitting mode and summarizing the different situations，the target resources were further reduced by local operations. The numerical experiments show that，compared with the results of solving the problem of non-split activities with resource window and the basic problem，the average value of the target for the 10，16，30，60，90 activities is 4.3%. For the comparison between the results of solving this problem and the non-split problem，the average optimization rate is 3.5%，which proves the effectiveness of the algorithm. At the same time，it is proved that the activity splitting is included in the Resource Investment Problem considering the resource window，which can improve the flexibility of problem solving and obtain better scheduling results.

Abstract:In order to investigate the normal force behavior of Magnetorheological Grease（MRG），three magnetorheological greases with carbonyl iron powder mass fractions of 30%，50% and 70% respectively were fabricated and measured by using a rotational rheometer in the mode of oscillatory shear. And the effects of magnetic field，time，strain amplitude，frequency and temperature on the normal force were systematically analyzed. The results show that the normal force values of three MRG samples increase with the increase of magnetic field strength. When the magnetic field strength is 740 kA/m，the maximum value of each sample reaches 6.97 N，8.93 N，14.91 N. Under different magnetic field levels，the influence of time on the normal force of magnetorheological grease undergoes three stages of slight decrease，constant，and slowly increasing. In the entire frequency range，the normal force value of the MRG with different magnetic field strengths changes slightly and basically at a stable value. There is a strain critical value，and the normal force increases with the strain amplitude at a faster speed before the critical value，further increases the strain amplitude and decreases the growth rate after exceeding the critical value. In addition，the normal force of the three MRG samples increases with increasing temperature and the MRG-70 normal force value is the largest. The above study of normal force in MRG is the theoretical basis for the design，optimization and application of magnetorheological grease devices.

Abstract:In order to improve the transmission capacity and the dynamic capacity increase of transmission line, the radial temperature rise phenomenon of transmission line is studied and analyzed. In this paper, the overhead transmission line is taken as an example. Firstly, the theoretical value of the conductor temperature is calculated based on the heat balance equation. Then, the electromagnetic coupling finite element stranding model of the transmission line temperature field is established to calculate the radial temperature distribution of the transmission line and study the influence of different factors on it. Finally, the effect of dynamic capacity expansion of the transmission line is analyzed based on the analysis results of the temperature field, and the minimum current carrying capacity is calculated according to the temperature distribution of the conductor. The results show that the radial temperature distribution of the transmission line is not uniform. The internal temperature is high and the surface temperature is low; the temperature of the transmission line is affected by different current, wind speed, ambient temperature and time. The radial temperature difference can generally reach 0.58~4.53℃, so the radial temperature difference of the overhead conductor is studied. According to the temperature distribution of the wire, the minimum current carrying capacity is calculated, which is beneficial to improve the dynamic capacity of the transmission line so as to ensure the safe operation of the line.

Abstract:Based on the potential advantages of unique mechanical performance and the micro-structure better design ability for cellular materials，this paper proposes a multi-segment energy absorption composite model filled with triangular and hexagonal honeycomb. Then, the dynamic response characteristic and the specific energy absorption of this model are numerically investigated by using explicit dynamic finite element (EDFE) method. The effects of honeycomb structure arrangement and relative density on the deformation mode, dynamic plateau stress, crushing load uniformity, and energy absorption capacity of the composite honeycombs are discussed in detail under different constant impact velocities. Research results show that the multi-segment filled honeycombs can realize the complementary advantages of type I and type II structures, which enables the axial force and bending deformation to participate in the overall deformation. Through the proper choice of cell micro-structures in each segment and the segment length, the crushing load efficiency of composite honeycomb is obviously improved and the fluctuation range of impact stress is significantly reduced. Composite honeycomb can effectively improve and control its energy absorption efficiency. These results are useful for the crashworthiness design and energy absorption controllable properties of cellular materials.

Abstract:Under the framework of crystal plasticity theory, a cyclic constitutive model based on dislocation density for face-centered cubic metals is proposed. The total dislocations are discretized into edge and screw components, and the multiplication, annihilation and interaction of dislocations are considered as the basic evolutionary mechanisms. At the same time, a cyclic constitutive model of single crystal is established by using the modified non-linear kinematic hardening rule. Then, the model is extended from single crystal scale to polycrystalline scale by explicit scale transition rule. The ratchetting strain of polycrystalline copper with typical face-centered cubic structure is simulated by using the proposed model. The numerical results show that the model can not only simulate the ratchetting strain and cyclic hardening characteristics of materials at polycrystalline scale, but also predict the ratchetting of materials at different orientations and stress levels from single crystal scale.

Abstract:In order to solve the problem of tracking failure caused by complex scenarios such as fast motion, occlusion and scale variation, a high-speed correlation filtering target tracking algorithm integrating motion state information is proposed. This paper makes three improvements based on the traditional Discriminative Correlation Filter: (1) The Kalman filter is added to the tracking process to modify the predicted position by using the motion state information, so as to deal with the tracking failure caused by fast motion and improve the tracking accuracy; (2) A separate filter for scale estimation is learned and the PCA method for dimension reduction of features is used to improve the tracking speed. (3) A high-confidence update strategy is proposed to determine whether the position filter is updated and whether the predicted position is transferred to Kalman filter for correction. The algorithm is tested on OTB-100 platform with several state-of-the-art tracking algorithms. Experiments show that our algorithm's average precision and success rate can reach 74.8% and 69.8%, respectively, and the average speed is 84.37 frames per second. Compared with other algorithms, the proposed algorithm can effectively improve the tracking performance, guarantee the tracking speed, and keep good tracking effect under complex conditions such as occlusion, ambiguous background and fast motion.

Abstract:Due to the particularity and danger of coal production, safety accidents often occur in the coal production process. Human factors account for a very high proportion. Therefore, it is necessary to study the violations of mine workers. Aiming at the singular point and time complexity problems often found in traditional dynamic time warping algorithms in human behavior recognition, a piecewise linear approximation algorithm combined with adaptive weight dynamic time warping algorithm is proposed. Then, the algorithm is simulated and experimented. The average recognition rate of the algorithm in the SDU Fall Dataset data set is 95.33%, and the average recognition time is reduced by 46.47%. Finally, we use the system to test in the coal mine. The results show that the proposed algorithm has a certain degree of improvement in recognition speed and accuracy.

Abstract:There are a large number of complex polygons containing thousands of holes (or even nested holes) in the land cover/land use vector data, and the existing spatial data indexing method has failed to indicate the inclusion relationship between complex polygons and their holes, resulting in computationally heavy and inefficient processing such as spatial data conflict detection and updating. In order to solve this problem, an improved quadtree spatial index method with inclusion relations of the complex polygons is presented in this paper. The method classifies the polygons in the nodes into five types according to the way they intersect the axes in the corresponding quadrant of the quadtree, i.e., intersect only the X positive axis, intersect only the X negative axis, intersect only the Y positive axis, intersect only the Y negative axis, and intersect both X and Y axes, and stores each of these polygons in five sublists (buckets) in the corresponding hierarchical index nodes, and then stores the parent-child inclusion relationship between the polygons in the node polygon objects. The authors developed the spatial index structure with inclusion relations and the algorithms of the corresponding operations（e.g.，insert, delete and query）for the complex polygons. The effectiveness of the approach in this paper is verified by an experiment of land cover data incremental updating, experimental results show that the time efficiency of the incremental updating is increased about several times using the proposed index method than that of the traditional quadtree index, and the improvement in efficiency is more significant with increasing data volume.

Abstract:Traditional ant colony algorithm is prone to cause failure by deadlocks in complex environments. A novel method is proposed to solve the problem，which adjusts the heuristic function by introducing environmental factors according to environmental information，increased the number of ants effectively，improved the search speed of ant colony，and expanded the search range. Aiming at the limitation of traditional ant colony algorithm in pursuit of shortest path in the ideal region in path planning，and the shortest path in the multi-factor environment is often not the optimal solution，the multi-objective path planning is proposed based on the weighted optimization of transition probability in different environments on the basis of the shortest path，which enriches the practicality and practical significance of the ant colony algorithm. Finally，the simulation experiment of optimization algorithm proves the feasibility of the method.

Abstract:Aiming at the shortage of detection range of substation metering device in the prior art，a new detection system scheme is proposed. The design divides the overall system solution into the underlying device layer，detection layer，data processing layer，and data analysis application layer，thereby realizing remote data transmission and remote data monitoring of the underlying substation metering device detection. The Life Cycle Management （LCC） module is designed. The LCC module is designed according to the relationship between safety，efficiency and cycle cost. The state of the substation metering device is evaluated by the random forest algorithm，and the life cycle of the asset is realized. Manage effective needs in terms of information acquisition and status assessment. Experiments show that the technical solution designed in this paper has less error.

Abstract:The identification and analysis of seismic gaps is one of the important means of medium-term earthquake prediction. However, it is difficult to achieve the desired effect by the traditional artificial method. Computer vision provides a new way to solve the problem. In this paper, an automatic identification method for seismic gaps based on image processing methods is proposed. The input is the text information of the historical earthquake. It is processed by computer vision methods such as distance transformation, threshold segmentation and watershed algorithm. The effective seismic gap is screened by iterative comparison and feature parameters. The output is the distribution image of the seismic gaps and its corresponding characteristic parameters. In addition, the algorithm of this paper is tested through a certain case. The test suggests that the algorithm of this paper can clearly obtain the seismic gaps with internal connectivity and clear external contour. Compared with the expert calibration, the recall rate of this algorithm is 81.25%, and the accuracy is 92.86%. This method provides a powerful tool for seismic researchers to conduct earthquake prediction business and related research.

Abstract:Aiming at the shortcomings of convolutional neural network in the recognition of traffic signs that the real-time performance is not good，and the equipment hardware requirements are too high，a real-time and high-precision improved network based on lightweight convolutional neural network is proposed. Separate convolution and activation function Mish，speed up the network training and recognition speed，reduce the requirements for hardware equipment；on the other hand，through the improvement of the network architecture and level，while reasonably changing the size and number of convolution kernels，the expression of image features and transfer. The experimental results on the BelgiumTSC traffic sign dataset show that the improved network significantly increases the network training speed，and the recognition accuracy is slightly higher than the original network，which verifies the effectiveness of the method in this paper. Compared with other models，this model can complete the task of traffic sign recognition more quickly and accurately，which verifies the feasibility of this method.

Abstract:The words of National Social Science Foundation (NSSF) titles are expressed into the train and test corpus. And then, the category determination model of the NSSF project by using the conditional random field model and the bidirectional short and long time memory model is verified from many angles and levels. The results are compared with the experimental results of the support vector machine model. Based on the corresponding model performance evaluation indexes, this paper not only verifies the overall performance of the traditional machine learning model on the small-scale corpus, but also proves that the overall performance of the conditional random field model with the artificial feature model is not certain to be outstanding, meanwhile, the performance of the conditional random field model is analyzed in a case.

Abstract:Color error represents real color reproduction ability on survelliance image within different color temperature. However，testing method in lab is reference-based and high demanded，which can not be applied for real image survelliance system. An illumination estimation of color check-check chart is obtained in lab. Meanwhile，color error value also is computed. A machine-learning method is used for compute the training model for illumination and color error value. Finally，we get color error value for real surveillance image through image illumination estimation and trained model.

Abstract:Recognizing human actions in videos quickly and effectively，has broad application prospects and potential economic value. Deep learning has been widely used for action recognition. We proposed self-attention temporal segment networks，whose inputs are clipped video clips. This network is based on deep networks and non-local means. By adding non-local modules to temporal segment networks with ResNet as the basic model，we can get our new model. Verified on TDAP dataset，our new model can recognize human actions more accurately than the original model，without increasing much time complexity.

Abstract:Humic acids （HAs） with different molecular weights were obtained by ultrafiltration method，and their molecular structures were respectively characterized by FTIR，UV-vis and Fluorescence spectra. A novel electro-Fenton process based on Pd/Fe3O4 nanocatalyst was applied for the effective and simultaneous removal of HAs and Cr（VI）. The results showed that a better mineralization efficiency of HAs was achieved with the increase of their molecular weights，following a descending trend of HA5（87.6%） > HA4（79.0%） > HA3（76.8%） > HA2（70.0%） > HA1（62.9%）. The HAs with high molecular weights was found to exhibit higher removal efficiency，which was attributed to the electrostatic interaction between the negatively charged HAs molecules and the Pd/Fe3O4 nanoparticles. Moreover，the HAs with high molecular weights that possess more conjugated structures and more substituted groups on the benzene rings were readily attacked by the reactive species ·OH，underwent ring-opening reactions till completely mineralized into CO2，whereas the HAs with low molecular weights that have high percentage of carboxylic acids were less active to react with ·OH，resulting in a weak TOC removal efficiency. Meanwhile，the removal of total Cr followed the same decreasing trend of HA5（91.8%） > HA4（88.2%） > HA3（85.9%） > HA2（85.4%） > HA（85.1%），for the HAs with high molecular weights could provide more complexation sites to chromium species，contributing to the removal of Cr（VI） in the developed electrochemical system. In this work，we show that the molecular structures of HAs have effects on the catalytic performance of Pd/Fe3O4 nanoparticles-promoted electrochemical process for the removal of TOC and Cr（VI）.

Abstract:With the further exploration of the ocean by human beings，the ocean has occupied an increasingly important position in the development of the country. How to comprehensively apply the sky-based means to obtain marine information and realize multi-base collaborative accomplishment of the maritime target mission has become particularly important. The system mainly combines space-based and space-based，and realizes the task planning of sky integrated ocean remote sensing resources to complete the identification and monitoring of maritime wide-area targets and moving targets. After the simulation experiment，the sky-based collaborative maritime target mission planning is completed，and multi-source remote sensing fusion is realized.

Abstract:By selecting the red paddy soil，taking the chitosan (modified)-zeolite (N-methylene chitosan hypophosphate-zeolite (ZHC), N-methylene chitosan phosphite-zeolite (ZPC)) as passivation agent and considering the factors of different processing time and amount of passivator added, the effect of chitosan (modified)-zeolite on passivation of Cd in soil was compared and analyzed. The results showed that pH value of soil could effectively increase by using the both passivants. The content of Cd in acid extractable in paddy soil reduced obviously. The passivating effect of passivating agent increased significantly with the passage of time. By comparing different types and different doses of passivation agents, the best passivation effect was that the additive amount was 3.2% ZHC. Using toxicity characteristic leaching procedure (TCLP) extraction, the concentration of heavy metals in red paddy soil decreased by 48.9%.

Abstract:Aiming at the problems in the actual two-way referral process of medical alliance platform，such as untimely referral arrangement and unsatisfactory patients，a two-way referral intelligent decision method based on random forest was designed. In this method，five main factors affecting referral were selected and an intelligent referral scoring model was established，including the number of beds，bed utilization rate，disease cure rate，treatment cost and distance. Then，big data and random forest methods are used to analyze the main factors and measure the suitability of the hospital to be referred to in the form of scores. The practical results show that predictions based on these five main factors can list a variety of referral plans，provide a basis for accurate decision-making on referral arrangements，and improve the efficiency of referrals.

Abstract:With the development of smart grid technology，power marketing management has become the core business of current development，and the quality of its work is directly proportional to the economic benefits of power supply companies. In order to effectively save energy and reduce load consumption in the power grid system，this paper proposes a new power marketing management system optimization program，which divides the power marketing management system management platform into marketing decision-making layer，marketing work quality management layer，marketing business layer and The customer service layer manages different types of power information in different categories. And through the use of electricity optimization models and algorithms，the load of the electricity price is relatively high during the period when the electricity price is relatively low，and the power utilization effect is improved，and the power consumption of the grid load is reduced. Tests show that this technical solution has better optimization performance and powerfully promotes the development of new technologies in the power marketing management system.

Abstract:Aiming at the energy consumption of ultrafiltration circulating pump in landfill leachate treatment，this paper designs a method and system to reduce the energy consumption of ultrafiltration circulating pump. The system adopts the B/S architecture and combines the optimal algorithm model trained by the neural network to adjust the operating parameters of the ultrafiltration circulating pump in advance，thereby saving energy and reducing consumption. The front-end display part uses Jquery+Bootstrap+Highcharts+ACE framework，the back-end service part adopts Spring+SpringBoot+MyBatis framework，and the front and back ends interact through Json data format. After the online test and on-site trial operation，the system has been officially put into use and achieved good results.

Abstract:Due to the lack of effective management means，the integrity，consistency and validity of power grid communication resources data can not be guaranteed，which affects the quality of service of power grid communication profession. To solve this problem，a metadata-based data verification method for power network communication resources is proposed. The method is divided into three processes：using network packet capturing to get the data of power grid communication resources，storing the captured data into disk array and multi-level verification of power grid communication resources data. Information layer verification in multi-level verification uses XSD recommendation standard as metadata driver and XSD verification engine to realize information verification. Model layer verification uses ontology files described by metadata model and public information model OWL as metadata to verify RDF format model extracted from the resource information of each system in nuclear power network communication network. Examples：Data layer checking is based on the metadata of the topological rules in the metadata model to check the common topological structure problems such as islands，inner and outer rings in various systems of nuclear power network. The experimental results show that the proposed method can effectively verify the consistency，validity and integrity of power grid communication resource data，and the verification accuracy is higher than 95%.

Abstract:In order to improve the production scheduling ability of the cigarette，a high-efficiency cigarette production scheduling method based on the support of the MES system is put forward. The MES system support model is constructed，and the output value of the production schedule is calculated in the model. And taking the scheduling output value as an adaptive learning coefficient，establishing a quantitative evaluation parameter screening model of the cigarette production scheduling，screening the quantitative evaluation parameter of the cigarette production scheduling，and outputting the test statistic quantity of the cigarette production scheduling，and constructing an optimized dispatching function of the cigarette production so as to realize the optimization of the cigarette production scheduling. The simulation results show that the method is good in convergence and high in yield，and is suitable for cigarette production scheduling.

Abstract:in view of the large prediction error of the traditional prediction model of the number of overseas tourists，a RBF neural network prediction model of the number of overseas tourists is designed. Firstly，the tourism sample data is normalized，and the number of overseas tourists is counted. Then，the prediction model of overseas tourists based on RBF neural network is designed through five processes：prediction population construction，evaluation of fitness value，penalty item setting，prediction sequence stability test and model prediction. Finally，experiments show that the RBF neural network model designed in this paper has less error than the traditional model，and can accurately predict the number of overseas tourists.

Abstract:In the current distribution-driven IoT data-driven business，the distribution of data resources is poorly balanced and planning efficiency is low，which affects the realization of collaborative functions. This paper improves the data-driven business of distribution IoT and the collaborative function of edge cloud. Event-driven collection of bottom-level electricity data，pre-processing to generate interval data，and unified data control to achieve data-driven business. The API technology is used to keep the computer in communication with devices with socket interfaces，the communication function between the cloud computing center and the edge devices is designed，and the control module function in the distribution network is set up，so as to realize the collaborative function of the edge cloud and ensure the global operation of the distribution network. The test results show that：under the same test conditions，compared with the traditional distribution IoT data-driven business and collaborative functions，the proposed distribution IoT data-driven business has a higher balance of data resource allocation and effectively improves the distribution of electricity The efficiency of connected data resource planning ensures the normal operation of the edge cloud collaboration function and meets actual application needs.

Abstract:Network traffic prediction is one of the effective way to improve user QoS. The network topology information is not fully utilized in current network algorithm prediction. A network traffic detection model based on high order graph convolutional network algorithm is proposed，and further predicts network congestion based on traffic information. The traffic prediction model utilizes the graph convolutional to capture the mix-hop effect of traffic. And the gated recurrent unit （GRU） obtains the time correlation information of the traffic in the network. The autoencoder model implements the unsupervised learning and traffic prediction. The simulation experiment is on the real data of the network Abilene. The experimental results show that the mean absolute percentage error（MAPE） value of the method in network traffic detection is 41.56%，which is lower 1.64% than DCRNN methods，at the same time，the prediction accuracy is also optimal.

Abstract:In order to realize the optimal management of power customers in the smart power management system，it is necessary to construct a stereoscopic image of the power customer，and propose a power customer stereoscopic image construction method based on multi-dimensional category feature recognition and corner point identification. The user portrait information tracking acquisition model uses the spatial feature domain classification method to classify the user image information，and uses the multi-scale layer-by-layer analysis method to accurately locate the fuzzy power customer image，and extracts the power category feature quantity of the customer stereo image，using the feature domain. The classification and block matching method performs error repair of the stereo image to realize fast and accurate positioning of the features of the user image. The support demand vector machine learning algorithm is used to adaptively classify the extracted power demand feature of the power customer stereo image，and realize the multi-dimensional construction of the power customer stereo image in the smart power management system. The results show that the method of constructing the power customer stereo image has better feature segmentation ability and the customer stereo image information has high accuracy.

Abstract:A software-defined network （SDN） widely used in hospitals is difficult to deal with internal network threats. A Bayesian-based trust management mechanism is designed to identify possible malicious devices inside the network. The mechanism mainly uses the Bayesian inference method to derive the probability of sending malicious attack packets to realize the trust management of the internal devices of the network. Experiments in the simulation environment and the real network environment prove that the method can reduce the trust value of malicious devices faster than similar methods..