Abstract:Anti-virtualization is currently an important factor affecting the overall acquisition of sample behaviour data by a dynamic analysis system of malicious code. This study proposes a systematic anti-virtualization confrontation method from host environment, network environment, and user interaction environment of dynamic analysis environment of malicious code, and implements the anti-virtualization confrontation in the existing dynamic analysis system. Experimental results show that the anti-virtualization confrontation effectively enhances the dynamic analysis system's ability to capture sample behavior data.

Abstract:In order to solve the security problems such as potential tampering and forgery in the traditional supply chain of transaction data, this study proposes a new trusted data management scheme based on blockchain technology. First, an intelligent contract suitable for supply chain is designed to code trade rules in order to prevent performance risks and improve the credibility of trade data processing. Then, ZSS04 scheme and sampling techniques are used to complete transaction data integrity check. Furthermore, a distributed consensus mechanism for supply chain is designed to improve the reliability of trade data storage. Finally, the trusted data management in untrusted environment is realized by using the native characteristics of blockchain. Results show that the scheme can provide new ideas and technical support for transaction data management in supply chain.

Abstract:Vertical search engine has always been a hotspot in the study of searching technique. Dispite a wide range of applications, the mainstream method of vertical search engine still has several flaws. In many cases, only a few stages have been optimized in the construction process of vertical search engine. Also, when obtaining information from websites, most of the methods require manual configuration, which is cumbersome. Based on an in-depth study of the vertical search engine technology, this article presents a method that uses JAVA open source tools such as Heritrix, Solr, combined with the extraction algorithm of web content and integrity word for automatically constructing a vertical search engine. In addition, the article examines the key issues in the various stages of the method's implementation and puts forward the corresponding optimization plan, which are examined to have strong practicality.

Abstract:With the development of artificial intelligence, scene recognition has attracted more and more researchers' attention, which is one of the important directions of computer vision research. The traditional manual features cannot sufficiently describe the characteristics of the scene images, which leading to unsatisfied performance. On the contrary, the features extracted from Convolutional Neural Networks (CNN) contain rich semantics and structural information of the scene images. As one of the most common architectures, AlexNet network model is chosen in this study. By improving the following 4 aspects of the network:depth, width,multi-scale extraction, and multilayer fusion, the proposed approach achieves high accuracies of 92.0% and 94.5% on two publicly available datasets respectively, showing the superiority compared with other methods.

Abstract:At present, most of the protection technology based on location privacy is designed for the user to carry out a single LBS request, it only protects the location of the current real user, but ignores the situation where the real user location is leaked by the cooperative user's overlapping when the real user is repeatedly queried. In this scenario, location prediction based on the real user position is used by the attacker to track the real user trajectories, resulting in the leakage of real user location privacy. In this study, a Location Privacy Protection Method Based on Similar Path (LPBSP) is proposed when the user initiates a continuous LBS request. Firstly, a certain equilibrium process is carried out through the history user density in the grid structure to make it conform to the real environment conditions, and then the similar path constructed in the adjacent time is carried out. The trajectory offset and speed similarity are constrained to make it closer to the real users, so as to confuse the attackers and achieve the purpose of location privacy protection. Finally, the feasibility of anonymous success rate, execution time, and location privacy protection is verified by simulation experiments.

Abstract:To address the needs of deep-buried oil and gas pipeline inspection and supervision, as well as the problems of low efficiency, poor timeliness, and low safety of conventional manual inspections, we design and develop a UAV oil/gas pipeline inspection system which combines UAV, convolutional neural network algorithms, and computer system integration technologies. Firstly, we introduce the overall design plan and operation flow of the patrol inspection system. Secondly, we present the system components. The system consists of four subsystems:UAV flight platform, neural network target detection system, UAV inspection management system, and enforcement terminals. The UAV flight platform uses oil-moving fixed-wing UAVs as the flight carrier, carries high-definition cameras for data acquisition, and the neural network target detection system automatically reads the image data, to detect, identify, and search the hidden dangers of engineering vehicles and pipelines along the route. The UAV inspection management system realizes the storage, management, and distribution of data information. The enforcement terminals receive hidden target information and perform rapid on-site enforcement. Finally, the application of the system and the subsequent development direction are summarized and forecasted. The system has been successfully applied to oil and gas pipeline inspection and supervision operations in Henan, Gansu, and other provinces. The results show that the system meets the field needs of oil and gas pipeline inspection and supervision.

Abstract:Aiming at the problem that the environmental regulatory authorities have difficulty in obtaining evidence from the excessive emissions of pollutant discharge enterprises, this study proposes a new method of online remote video forensics system of environmental protection. The method enables the environmental protection departments to remotely monitor the pollution discharge of polluting enterprises in real time and obtain evidence from history videos by means of streaming media distribution technology, XML-RPC protocol authentication technology, as well as the audio and video flow optimization technique in the ADSL network environment. The web video controls developed in this project combined with the pollution sources online monitoring system form a comprehensive online remote video forensics system of environmental protection. Thus, a convenient platform is provided for environmental protection and supervision departments to better regulate those pollution-discharge enterprises and execute forensics.

Abstract:As Location-Based Social Network (LBSN) services become increasingly popular, next Point-Of-Interests (POI) recommendation emerges as one of many important applications of LBSNs. With the growing ability of collecting information, more and more temporal, spatial, social contextual and semantic tags information is collected in systems, which makes the location prediction problem becomes feasible. Some works, like Factorizing Personalized Markov Chain (FPMC), Tensor Factorization (TF), Recurrent Neural Networks (RNN), etc., have been proposed to address this problem, but they all have their limitations. In this study, we extend Long-Short memory recurrent neural networks (LSTM) and propose a novel method called POI-LSTM. POI-LSTM can model social contextual and semantic tags information in each layer, and employ temporal and spatial contexts in more efficient way. Experimental results show that the proposed POI-LSTM model yields significant improvements over the competitive compared methods on two typical datasets, i.e., Yelp and Foursquare dataset.

Abstract:With the rapid development of smart grid, the growth of power system data is also very fast. There is an urgent need of in-depth research for power big data. Because of the large span of the data acquistion speed, numerous data sources, complicated interaction interfaces, and various kinds of data type, the existing big data technologies are unable to adapt to power big data aquisition. In this study, a new solution for power big data acquisition is proposed. The system schedules and manages the heterogeneous data source acquisition tasks through mixed data acquisition model and collection cluster. With the technology of data confidential degree label, the system preserves the original data, indicates the quality of the data and provides convenience for big data analysis applications. The system submittes collected data to the big data platform for storage by Sqoop, Kafka, file transfer, or other methods. The system has been deployed and puts into use in the user site. It runs stably and has a sound effect.

Abstract:With the acceleration of population aging and the increasing of sub-health population, the demand of health monitoring service is expanding. Heart rate, blood pressure, blood glucose, and other physiological data contain important pathological information of human health status. In order to receive and preprocess these physiological data, and complete the centralized management of physiological detection equipment and automatic data collection, a family health monitoring gateway is developed based on the Raspberry Pi research. The family health monitoring gateway uses Bluetooth communications to obtain data from physiological monitoring devices, and stores or uses wireless networks to transmit it to the server. Finally, taking the monitoring object's electrocardio data acquisition, transmission, and display process as an example, the whole system is tested. The gateway provides a feasible scheme for family health monitoring and is of great significance in alleviating social medical pressure.

Abstract:Prediction of future behavior of complex set of data sets is a difficult task. Data mining is a potential technical way. For the real-time evolutionary data sets containing multiple time series and non time sequence, a method of integrating the sequence segmentation, clustering, and pattern matching is proposed, which combines the theme discovery and joint decision. In the whole method construction, the topic discovery prediction and joint decision prediction are fused into the early sequence segmentation and clustering. The sequences are stratified and segmented for forming standard pattern sets of each layer, using multi time granularity and multi span. Then, according to the standard pattern set, with the prediction strategy, the compound pattern with high stability extension behavior is used as the theme pattern. This can predict with online pattern matching. Finally, a distributed parallel computing architecture is used to implement the whole processing algorithm. Theoretical deduction and experimental data analysis show that the accuracy of the method is improved compared with the traditional time series prediction method.

Abstract:Underwater vehicle requires small volume, stable motion, low power consumption, reliable performance, and easy operation. In this study, the motion control system of underwater vehicle is built with STM32F407 as the main control unit, and the software structure and data acquisition flow are designed. The data test of the thruster is carried out, and the ROV space motion coordinate system is established. The equations of motion of the vertical plane are obtained and the simulation analysis of the angle and depth of the vertical plane under the step response is carried out. The stability and reliability of the motion control system are further verified.

Abstract:As application scenarios become more and more complex, heterogeneous multi-core architecture becomes the mainstream architecture of embedded processors. At present, there are many limitations in the application of the single operating system model mainly used in multi-core processors. In order to give full play to the heterogeneity of multi-core processor features, it is essential to deploy the corresponding operating system for different cores of heterogeneous processors and realize the collaborative processing technology of multiple operating systems. In this study, the operating system of heterogeneous multi-core processor (ARM+DSP) is investigated, and embedded Linux and ReWorks of domestic DSP real-time operating system are successfully transplanted on heterogeneous multi-core platform. In order to realize the collaborative processing between ReWorks and Linux operating system, this study analyzes the key technologies of inter-core communication, and takes AM5718 of TI as an example, designs a series of multi-core heterogeneous communication components. It has been tested that the heterogeneous communication components can dynamically load the DSP kernel with the ReWorks operating system and applications, send and receive messages between Linux and ReWorks cores, and perform collaborative computing between Linux and ReWorks.

Abstract:With the continuous expansion of business in meteorological system, the performance, stability, and maintenance efficiency of original core network have been short of capacity to meet the demand. To guarantee reliable and efficient operation of meteorological network system, current status of the network has been studied and rebuild demands are proposed, network reform schemes and equipment changeover procedures have been designed in detail, a new generation of CSS and iStack switch stacks technologies have been adopted in smoothly upgrading the meteorological core network based on network virtualization and visual fusion operation and maintenance platform, overall built a 2 layer flat meteorological network system. By rebuilding, the network structure is simplified, and the network performance and maintenance efficiency are improved.

Abstract:The current reefer container ship controls the refrigerated containers individually, such mechanism lacks of unified dispatch management of the refrigerators. The power demand of single refrigerator is random, resulting in large peak-valley difference of the total electric power demand, which further affects the power allocation and efficiency of ship power station. In order to solve the above problems, a reasonable dispatch of the refrigerated containers should be carried out with the prerequisite of ensuring the safety of temperature. This study proposes a scheduling algorithm based on quantum genetic algorithm for power balancing to find the optimal scheduling strategy for refrigerated containers. Firstly, this study establishes a mathematical model for the optimal scheduling of refrigerated containers, determining the optimization targets and constraint conditions. Secondly, it uses the Genetic Algorithm (GA) and Quantum GA (QGA) to solve the objective function, followed the comparison of their actual power changes before and after the scheduling and evaluation of the optimal scheduling capability of the two algorithms. The experimental results show that both QGA and GA can realize the optimal scheduling of refrigerated containers and reduce the peak-valley difference of total power demand, thus balance the power load. Nevertheless, QGA converges faster than GA, and its ability is stronger than that of QA in terms of balancing power demand and optimizing power station.

Abstract:Aiming at the problem of high computational complexity of human pose estimation algorithm in deep learning field, a fast human pose estimation algorithm based on optical flow is proposed. Based on the original algorithm, using the time correlation between video frames, the original video sequence is divided into key frames and non-key frames, which are processed respectively (the images between two adjacent key frames and the forward key frame compose a video frame group, which is similar to the frames in the same video frame group), the human pose estimation algorithm is applied only to the key frames, and the key frame recognition result is propagated to other non-key frames through the lightweight optical flow field. Secondly, aiming at the dynamic characteristics of the video field, this study proposes an adaptive key frame detection algorithm based on local optical flow to determine the position of the key frame of video according to the local time-domain characteristics of the video. The experimental results in OutdoorPose and HumanEvaI data sets show that the detection performance of the proposed algorithm is slightly higher than the original algorithm in the video sequences with complex background and component occlusion. The detection speed is increased by 89.6% in average.

Abstract:In the marine remote sensing image data processing and the characteristics of visual detection in the process of marine information, mesoscale vortex information detection is one category of the important content. Its results directly affect the accuracy of marine mesoscale vortex information retrieval. In order to improve the accuracy of the mesoscale vortex information extraction, the remote sensing images are detected step by step based on the imaging characteristics of mesoscale vortex and combined the technology of connected region extraction. According to the characteristics of the marine mesoscale vortex remote images, the features of marine mesoscale vortex are detected based on edge detection and the connected region extraction technology. The closed contour of the vortex in the image and the characteristic parameters of spiral vortex are extracted based on the shape and the scale criterion of the vortex. By using Matlab and C# hybrid programming, the mesoscale vortex automatic detection and visualization software are realized by combining the mesoscale vortex detection algorithm and the remote sensing data information. This method can improve the efficiency of the remote sensing image processing. The experiment results show that this method can achieve high mesoscale vortex information detection precision and has preferable detection effect.

Abstract:Dynamic fault tree analysis is an important reliability analysis technique for complex systems. However, there are serious state space explosion problems in traditional modular methods such as binary decision diagrams. This paper systematically introduces a dynamic fault tree analysis method of edge-value decision diagram, in which the Edge-Value Multiple-valued Decision Diagram(EVMDD) has a more compact representation function than other existing decision diagrams. By reducing the number of states, the computation time is shortened and the state space explosion problem can be effectively alleviated. The example demonstrates the method and advantage of EVMDD for multi-state systems and multi-functional systems.

Abstract:The Correlation Interferometer Algorithm (CIA) is commonly used in the Direction Of Arrival (DOA) estimation. But when it realizes the direction finding, the long and short baselines in the array lead to the problems of baseline symmetry and phase ambiguity. An improved correlation interferometer algorithm based on quadrant classification is proposed. First, the time difference of signals arriving at each element is converted into phase difference, as well as the obtained phase difference is compared with 360°, and the obtained integer and remainder are recorded. Then the remainder is quadrant classified, after which the initial estimated value of the signal is obtained by using a traditional correlation interferometer algorithm. Finally, the final estimate value of the signal is calculated based on the inverse operation. Experimental simulations show that the improved algorithm successfully solves the problems of baseline mirror symmetry and phase ambiguity. Moreover, the accuracy of signal estimation is improved, the computational complexity is reduced, and the real-time performance of direction finding is improved. Therefore, it has significant value in the DOA estimation.

Abstract:The traditional moving target tracking and prediction algorithm is difficult to ensure that the robot can capture and predict the high-speed moving targets in advance. In particular, the collision of moving targets in the course of sliding changes the original motion direction. Aiming at this problem, a moving target tracking prediction algorithm based on frame difference and collision algorithm is proposed. Through the inter frame difference method, the specific position and velocity of the moving object in the plane are identified quickly, and the collision of the moving target is judged according to the direction of motion velocity. When the moving targets collide in the process of motion, the collision simulation model is established by using the LS-DYNA dynamic analysis software, and the collision algorithm is obtained by fitting the simulation data with MATLAB, and the motion trajectory of the moving target is predicted by the collision algorithm. The results show that the moving target detection and tracking algorithm, which combines the inter frame difference and the collision algorithm, is faster for the tracking and prediction of the moving target in the plane, and can fully meet the requirements of the robot's fast algorithm.

Abstract:The compressed-domain object tracking approaches utilize the information that is directly extracted from the compressed bitstream, such as motion vector and block coding modes. Because the existing compressed-domain tracking methods have poor tracking performance for small moving targets, this study proposes a compressed-domain tracking algorithm for small moving targets. By analyzing the shortages of the existing algorithms, the performance of small target tracking is improved from the acquisition of initial frame mask, the setting of outlier boundary and the edge control of the predicating small target, and some system parameters of the block-coding system are optimized through data-driven methodology. Experiential results on three small-target video sequences show that compared with other object tracking methods, the proposed method can effectively improve the tracking performance for small moving targets in terms of accuracy and F-measure.

Abstract:In recent years, the image super-resolution reconstruction has always been a hot research field, but the corresponding research results about arbitrary-scale-ratio super resolution are still rare. Under high scale ratio, the image resolution will become lower, and it is difficult for human eyes to recognize such image content. With the advancement of technology, machine vision has been used to recognize the images with very low resolution, and the research on arbitrary-scale-ratio super resolution has become increasingly important. Through testing various representative super-resolution algorithms, this study proposes a scale-adaptive super-resolution reconstruction algorithm according to a full-scale quality sum criterion after performing extensive arbitrary-scale-ratio analysis on image super resolution. Experimental results show that the proposed algorithm can achieve better overall-reconstruction performance within the whole scale range.

Abstract:For the robot trajectory planning problem, an improved Differential Evolution(DE) algorithm based on sliding mode control is proposed. The operation time and energy consumption is taken as the objective function. The direction of variation is guided by the difference between the best individuals and the average individuals in the population. Using the best individuals to replace the worst individuals in the population, the rate of convergence is accelerated. Removal of mutation factor and cross factor reduces manual intervention and enhances the stability of the model. Using piecewise Hermite interpolation of order 3 instead of spline interpolation of order 3 to prevent fitting overshoot and reduce chattering. Based on the system state space equation, a sliding mode control law was designed and stability is proved by using a Lyapunov function. Simulation experiments and the analysis of results are demonstrated that the improved algorithm not only has strong search ability but also accelerate the convergence rate and reduce the chattering effect in the trajectory planning of the system states.

Abstract:In order to solve the defect of the traditional spectral subtraction algorithm, an improved spectral subtraction based on the joint maximum a posteriori probability is proposed. The traditional spectral subtraction was used to reconstruct the speech via obtaining difference of the amplitude between the noisy speech and noise and extracting the phase of the noisy speech. "Music noise" was produced by the method, and the effect of signal enhancement under low signal-to-noise ratio was not ideal because of inaccurate phase estimation. For this, the multiband spectral subtraction and phase estimation were introduced, and spectral subtraction was carried out in the subbands which were obtained by spectrum division. And it has worked well on reducing the influence of "music noise". Meanwhile, the phase estimator based on the maximum a posteriori probability was constructed which was obtained by combining the amplitude function and thephase function of the signal and alternate iteration. The experimental results show that, compared with the traditional spectral subtraction, the proposed algorithm has performed better in terms of the quality perception and intelligibility of the enhanced speech at low signal to noise ratio.

Abstract:In order to help readers quickly grasp the main content of a large amount of news report information, this paper analyzes the impact of event elements on the main news content, and combines the basic principles and requirements of news reports, proposes a method of extracting event elements based on hybrid model. The proposed method first weighs the entities recognized in the news data, and then uses the dependency syntax tree to analyze the role of entities in news events, and dispels the reference phenomenon of elements. Finally, the fusion frequency and role relationship are used to improve the entity weighting method and effectively extract the important elements of news event relevance. The experimental results show that the method described in this study can accurately extract event elements with strong relevance to news events and improve the efficiency of readers' rapid selection of news event elements.

Abstract:The extraction of mineral particles from rock slice images is the basis for grain size analysis and mineral components recognition. In order to further improve the accuracy of mineral particles extraction, a new method of mineral particles segmentation is proposed. This method is based on the crossed polarization image sequences which are in the same view and different angles. After merging the sequence images, entropy rate superpixels algorithm is used to extract the target of mineral particles. In order to reduce the over segmentation of mineral particles, this study uses the fast region merging algorithm to merge regions with similarities. Finally, filtering over segmented regions and merging target particles again according to the change rules of mineral particles under crossed polarization to realize automatic segmentation of mineral particles. The proposed method can be used to extract mineral particles from rock slice images and achieve good results.

Abstract:As a real-time computing framework, Storm has provided an efficient, fast, and real-time processing ability for multi-source heterogeneous data processing. However, Storm's default scheduler uses a simple Round-robin method and unable to adjust assignments of its task according to cluster's dynamic load status. To solve this problem, this study proposes a load-balancing strategy based on performance-aware. It could calculate Performance-Aware Value (PAV) according to node's processing ability, then greedy scheduling to achieve load balancing, which assigns the amount of computation match with node current processing capacity to achieve load balancing. Compared with the default scheduling algorithm, the results show that this algorithm can effectively reduce the Storm processing delay and improve the throughput, finally achieve cluster's load balance.

Abstract:A higher intelligibility subspace speech-enhancement algorithm based on the relative Root Mean Square (RMS) of speech segmental Signal-to-Noise Ratio (SNR) with low SNR is proposed. Under harsh conditions of low SNR, an improvement of noisy speech quality based on the majority existing speech-enhancement algorithms is often accompanied by a decrease in speech intelligibility. One important reason is that these algorithms only use Minimum Mean Square Error (MMSE) to constrain speech distortions but ignore that speech distortions caused by speech enhancement algorithms have different intelligibility influences on different speech segments. In order to overcome this disadvantage, the RMS of short-time segmental SNR was used to classify speech segments. Then the gain matrix components of middle-level RMS segments were modified to reduce the influence of speech distortion on enhanced speech intelligibility. Objective evaluation shows that the improved algorithm can improve enhanced speech intelligibility Normalized Covariance Metric (NCM) evaluation values. Subjective audition shows that the proposed algorithm does improve the enhanced speech intelligibility.

Abstract:Feature matching is one of the key steps in image mosaic. The matching algorithm based on the best of two nearest matches often has a large number of mismatches. The good filtering algorithm can reduce the mismatch rate and improve the processing efficiency. Therefore, it is of great significance to study this kind of algorithm. The RANSAC algorithm is a widely used filtering algorithm, but it has many defects such as uncertain number of iterations and none of self-adaption in BA process. In this study, we propose a new filtering algorithm of Feature Matching based on Local Clustering (LCMF). The feature points are extracted by SURF and ORB, the BestOf2NearestMatcher algorithm is used to match, and then the LCMF algorithm is used to filter. The experiment shows that the algorithm can get better filtering result when ORB is used to extract feature.

Abstract:In order to improve the computation accuracy of the Muskingum flood routing model, a new method of parameter estimation for Muskingum model based on the improved grey wolf optimizer (IGWO) is proposed and applied to the flood calculation in the south canal between Chenggouwan and Linqing River. The experimental results show that IGWO can effectively estimate the parameters of the Muskingum model. Compared with the other parameter estimation methods of Muskingum routing model, IGWO has higher calculation accuracy and better optimization performance.

Abstract:This study considers comprehensively the various factors of grain production yield and optimizes primary BP neural network weights using the improved Particle Swarm Optimization (PSO) algorithm, then establishes a prediction model suitable for prediction of small sample grain yield. The experiment proves that this model has higher prediction precision and greater fitness than grain yield prediction model based on classical BP neural network and PSO-BP neural network.

Abstract:Aiming at the low utilization of storage space in traditional single-sparse storage model and the problem of more connections in reconstructing the logical relationship of tenants in the block table storage model, the combination of sparse tables and block table storage model is put forward. In the storage model, the attributes in the logic tables provided by the SaaS providers and the attributes in the custom logical tables are mapped into the corresponding sparse tables and the custom attributes of a part of the tenant common data type are stored in the block table, so as to avoid the problem of data migration caused by the number of extended columns exceeding the number of sparse tables. Finally, the query conversion and query efficiency are optimized through the query rewriter. The experimental results show that the storage model improves storage space utilization and query efficiency compared with the traditional sparse table storage model.

Abstract:Feature Selection (FS) is reducing dimensions and denoising. However, there are many factors that affect the features selection, mainly including the dimensions, importance, and semantic of terms. For feature representing high-dimensional but sparse of short text and traditional features extraction lack semantic, a feature selection function FS fusing multi-factors is constructed. It is verified that FS not only can integrate the semantics of the feature, but also can remove a large number of redundant features, thus improve the weight of the features with class distinction capabilities, comparing with the traditional feature selection function TF-IDF. FS as a new function, using the Chinese corpus of Sogou Lab for short text classification, verifys the effectiveness of the method.

Abstract:Distributed ICE middleware is an efficient RPC framework that supports cross language development, and the IceGrid service is the core of the whole framework. The main function is to provide services such as location management, remote management, load balancing and service registration for clients. In the distributed platform constructed by ICE, once the IceGrid service is in trouble, the whole network communication will have problems. IceGrid itself supports the high availability of the master slave mode, but when the master is not available, slave will not automatically upgrade to the master, and it must be restarted manually, which seriously affects the efficiency of the IceGrid. Zookeeper is an open source distributed application coordination service. It is a software that provides conformance services for distributed applications. This paper presents a highly available and improved model for IceGrid services based on Zookeeper. When the master is not accessible, Zookeeper selects a new master from one of slaves automatically.

Abstract:In order to find out more location information from users, the user feature is dug out from location semantic similarity. The location topic model is built for user sign-in information by using LDA algorithm, and distribution functions can be calculated by the Gibbs sampling algorithm. The user similarity feature vector based on sign-in location semantic is put forward by these distribution functions. Then, the supervised machine learning algorithm is put forward to make link prediction by multi-dimensional similarity feature vector from fusing LBSN network structure information, sign-in location information and location semantic similarity. The experiments result on Gowalla databases shows that the link prediction algorithm using more similarity feature as subsidiary information can improve performance of LBSN link prediction significantly comparing with the traditional algorithm.

Abstract:In this study, the gesture recognition based on SKIG RGB-D multimodal isolated gesture video is studied. The RGB and depth videos are extracted into the form of images. Then the sampled 32 frames from images are input to the densely connected 3DCNN component to learn short-term spatiotemporal features, after that the features input to the convolutional GRU to learn long-term spatiotemporal features. Finally, the trained networks for single modal are used to multimodal fusion to improve the recognition accuracy. 99.07% recognition accuracy is obtained on the SKIG dataset, which achieves high accuracy and proves the validity of the network model proposed in this study.

Abstract:IS-IS is a link state routing protocol, which is introduced into the TCP/IP protocol stack because of its high efficient and stable features from the OSI protocol stack. IS-IS supports both the CLNP and the IP routing. On the basis of the in-depth study of IS-IS protocol by NSAP address, area type, message analysis and working principle, an IS-IS experimental network model of multi-region and multi-network is designed, and it is successfully realized through the Dynamips simulation platform. The simulation results verify the mechanism and principle of IS-IS protocol with the analysis of packet capture and neighbor table, topology table, and LSDB, which can provide reference for deployment in the actual network, and the instance can be migrated to the actual network seamlessly.

Abstract:Existing decoding algorithms for polar codes still suffer from very high complexity. To solve this problem, an SCL decoding algorithm based on BP neural network is proposed. In offline, the algorithm builds and trains an appropriate BP neural network by collecting data. With the trained BP neural network, the optimal initial value of list size L is found through on-line operation. On this basis, the complexity is reduced by designing an improved SCL decoding algorithm. Experimental results show that compared with existing algorithms, the proposed algorithm can significantly reduce the average decoding complexity at low SNR.

Abstract:The statistical characteristics of taxi behavior have important significance to study the economic and psychological of human dynamics. Based on the taxi GPS track data in Xi'an, the lane-changing behavior was quantitatively studied by big data analysis technology. The model of a lane-changeing behavior recognition was designed, combined with the big data analysis technology, the number of taxi drivers' lane-changing was quantified in terms of different time periods, and the correlation analysis between taxi drivers' lane-changing times, taxi average driving speed, and taxi drivers' income was carried out. The results show that there is a significant negative correlation between the income of taxi drivers and the average driving speed of taxis, which further indicates that taxi drivers' habits and psychology have a significant impact on the whole taxi operation.

Abstract:The sea-sky background image has a large area of sky and the target must be near the sea-sky-line when looking from a distance. The existing dehazing algorithms improve the sky area by weakening the processing of the sky area. This will inevitably weaken the dehazing effect near the sea-sky-line and be detrimental to the next target detection. This study proposes the method of atmospheric scattering physical model to dehaze. Firstly, based on the features of the sea-sky background image, the edge detection algorithm is used to divide the image into sky and non-sky regions. Considering the physical meaning of atmospheric light, the maximum value of the sky region is estimated as the value of atmospheric light. Secondly, the cost function is designed based on the prior information that the fog image has low contrast and the fog-free image has high contrast, and dividing the image into blocks by SLIC superpixel segmentation. The rough transmission is estimated by finding the minimum of the function in each block, and then the guided filtering is used to eliminate block effect. Finally, the fog-free image can be obtained by substituting the parameters obtained in the first two steps into atmospheric scattering model. The experimental and analysis results show that this method can achieve better dehazing effect of sea-sky background images.

Abstract:In this paper, we present the study of network middleware of time series database, summarize and analyze the characteristics of common network middleware. According to the high performance requirement of time series database, the key technology of network middleware is studied. The concurrent model of multiple reactor and thread pool is adopted to improve the concurrency of thread, and the multi-frame message format is proposed to realize the dynamic expansion of the message protocol. Finally, the related network middleware is realized in the PCS-9000 time series database. The middleware achieves the expected performance requirements.

Abstract:Because of physiological monitoring data has time continuity, inaccuracy, and fuzziness, the traditional classification algorithm is difficult to be used directly. In view of the above problems, a classification method of chronic lesions based on rough sets is proposed. First, the physiological monitoring data are discretized based on fusion of correlation and Chi-merge statistics. Then, this method uses the attribute reduction algorithm based on the compatibility matrix to remove the redundant attributes of the data. Finally, classification rules are mined based on batch and incremental data, and intelligent classification of chronic diseases can be realized by applying the above rules based on MapReduce framework. Experiments show that the method has a high recognition rate, which is helpful for the individual to fully understand the health risks.

Abstract:The hardware functional modules of relay protection CPU board are analyzed. The automatic test system of CPU board has good universality and practicality, which is designed with the idea of modularization and hierarchy, and the hardware structure of the automatic test system is described in detail. According to the characteristics of CPU board hardware function modules, they are divided into three categories, and testing methods and testing processes of various functional modules are expounded respectively. Combined with a specific application case, the present application situation is introduced. It has achieved sound use effect in the production test of CPU board, and greatly shortens the test time in the production process of relay protection of CPU board.