15 June 2020, Volume 46 Issue 6
-
Select all|
-
Computer Engineering. 2020, 46(6): 1-11. https://doi.org/10.19678/j.issn.1000-3428.0057669Knowledge graph,as the development product of symbolism,is considered to be an important part of artificial intelligence technologies and systems,and widely used in encyclopedia knowledge,biological information,social network,network security and so on.Visualization query of knowledge graph is an important technology for understanding and analyzing knowledge graph,which can help ordinary users effectively query knowledge graph.This paper introduces the data models of knowledge graph and visualization technologies by type,and describes the general steps of large-scale knowledge graph visualization from the perspective of data scale.Then this paper analyzes visualization query languages based on RDF graph and property graph,keyword-based,filter-based and template-based visualization query systems,and visualization query method of ontology.This paper compares and summarizes the existing visualization query technologies of knowledge graph in terms of readability,learnability and user-friendliness.Meanwhile,the application of visualization query in domain-specific knowledge graph is also described.Finally,the future research directions of visualization query of knowledge graph are put forward as well.
-
Computer Engineering. 2020, 46(6): 12-19. https://doi.org/10.19678/j.issn.1000-3428.0056660To address the data security problems caused by vulnerabilities and backdoors in existing distributed storage systems,this paper proposes a feasible security construction method for system by introducing the Cyberspace Mimic Defense(CMD) theory and its related security mechanism.The architecture aims at enhancing the security protection ability of the system.During the design process,the main threats and attack ways to distribute storage systems are analyzed to locate system’s core weakness,and the cost and effectiveness of protection are also considered.Taking the Hadoop Distributed File System(HDFS) for big data as the target object,the mimic architecture for metadata services is designed.This paper builds the Dynamic Heterogeneous Redundancy(DHR) structure of metadata services to protect the core information and functions of the system.Then the heterogeneous placement of copies is implemented to protect user data.On the basis of this architecture,a collaborative arbitration and scheduling mechanism based on feedback information is proposed.Test results show that the proposed method can effectively improve the security of distributed storage system.
-
Computer Engineering. 2020, 46(6): 20-25. https://doi.org/10.19678/j.issn.1000-3428.0056274The main function of Mobile-Edge Computing Offloading(MECO),which is a key technology for mobile edge computing,is to migrate the compute-intensive tasks of Mobile Devices(MD) to edge servers to implement low-energy and low-latency services.However,transmission delay and energy consumption in offloading of compute tasks still affect user experience.To further reduce the delay and energy consumption,this paper proposes a power distribution algorithm based on game theory for MECO system.Under the constraints of computing resources of a server,the binary search method is used to optimize the transmission power to reduce transmission delay and energy consumption,and the non-cooperative game theory is used to solve the multi-user offloading decision problem in order to reduce the system overhead.Simulation results show that the proposed algorithm can obtain better computation offloading performance,which is increased by 41% and 12% respectively compared with the original game-based offloading algorithm and adaptive sequential game-based offloading algorithm.
-
Computer Engineering. 2020, 46(6): 26-33. https://doi.org/10.19678/j.issn.1000-3428.0056462In view of the problems existing in the detection of small vehicles in complex traffic scenes,this paper proposes an improved objection detection method based on YOLOv3,S-YOLOv3.The method uses ResNet to improve the feature extraction structure of Darknet-53 of YOLOv3.The four scales of features of the object are obtained by using the Feature Pyramid Network(FPN) to fuse shallow feature information and deep feature information.Then the impact weight of the loss function is adjusted according to the size of the detection target,so as to enhance the detection performance of small objects and occluded objects.Experimental results on KITTI dataset show that the average detection speed of S-YOLOv3 is increased to 52.45 frame/s and Mean Average Precision(mPA) is increased to 93.30%.Compared with the YOLOv3 method,this proposed method can improve the precision of small object detection while ensuring the real-time performance.
-
Computer Engineering. 2020, 46(6): 34-39. https://doi.org/10.19678/j.issn.1000-3428.0055520The incidence rate of traffic accidents increases in heavy rain due to the limited vision of drivers.In order to accurately predict the traffic flow of expressway under heavy rain and reduce accidents,this paper proposes a method based on the improved Cuckoo Search(CS) algorithm and Radial Basis Function(RBF) neural network for expressway traffic flow forecast under heavy rain.The method uses monkey climbing process in monkey swarm algorithm to optimize the cuckoo position update strategy,and then adopts the adaptive update strategy for recognition rate to establish a traffic flow forecast model for expressway based on improved CS-RBF Neural Network(CS-RBFNN).Experimental results show that the improved CS-RBFNN model has a higher convergence speed and prediction accuracy than the improved GSO-RBFNN model.The Mean Absolute Percentage Error(MAPE) of the proposed method is 8.2% and the Mean Absolute Error(MAE) is 20.14.The Root Mean Square Error(RMSE) of the method is 19.2,and its prediction accuracy is higher than 90%.
-
Computer Engineering. 2020, 46(6): 40-49. https://doi.org/10.19678/j.issn.1000-3428.0056559To achieve the effective acquisition of frame-level speech features under different emotional needs of children,an emotion recognition model for children speech based on improved Long Short-Time Memory(LSTM) network is established.Frame-level speech features are used to replace the traditional statistical features,so as to retain the time sequence relationship of the original speech.Introducing attention mechanism to convert the traditional forget gate and input gate into attention gate,and the deep attention gate is calculated according to the customized depth strategy,so as to improve the performance of speech emotion recognition.Experimental results show that compared with the traditional LSTM based recognition model,the recall rate and F1 score of the model are increased by 3.14%,5.50% and 1.84%,5.49% respectively on Fau Aibo children’s emotional data corpus and infant crying emotional demand database.Compared with the traditional LSTM and GRU based recognition model,the training time of the model is shorter and the recognition rate of children speech emotion is higher on CASIA Chinese emotion database.
-
Computer Engineering. 2020, 46(6): 50-59. https://doi.org/10.19678/j.issn.1000-3428.0054479Micromanipulation of Real-Time Strategy(RTS) game refers to manipulating multiple combat units to win in a game.Traditional search methods are inefficient and have limited search space in large-scale battle scenarios.To address the problem,this paper proposes a method combining deep learning with online search,using the learning model to guide the search process.A Joint Policy Network(JPN) based on encoding-decoding convolution architecture is given and embedded into three classic search methods:PGS,POE,and SSS+,so as to realize end-to-end learning of joint actions of multiple agents.Experimental results show that the method can adapt to complex combat scenarios,defeating the built-in artificial intelligence method in the two benchmark scenarios of StarCraft:BroodWar,and the winning rates are 95% and 99% respectively,which are close to the current best benchmark method.
-
Computer Engineering. 2020, 46(6): 60-64. https://doi.org/10.19678/j.issn.1000-3428.0054540This paper constructs a CGAtten-GRU model based on dual-encoder network structure to solve the problem that the encoder cannot fully encode the source text in the sequence-to-sequence(seq2seq) model.The two encoders use Convolutional Neural Network(CNN) and Bidirectional Gated Recurrent Unit(BiGRU) respectively,and the source text enters the two encoders in parallel.An attention mechanism is constructed by means of the outputs of two encoding networks.The decoder uses GRU network combining the Copy mechanism and the beam search method to improve the accuracy of decoding.Experimental results on large-scale Chinese short text summarization dataset LCSTS show that compared with the RNN context model,the proposed model improves Rouge-1 by 0.1,Rouge-2 by 0.059,and Rouge-L by 0.046.
-
Computer Engineering. 2020, 46(6): 65-72. https://doi.org/10.19678/j.issn.1000-3428.0054127In continuous dimensional emotion recognition,the part of highlighting emotional expression varies in each modality,and different modalities also have different influence on emotional states.To address the problem,by learning modal features and fusing them in a reasonable way,this paper proposes a multimodal dimensional emotion recognition model based on Hierarchical Attention Mechanism(HAM).Frequency attention mechanism is added to the audio modality to learn the context information in frequency domain,and the video features are fused with the audio features by using the multimodal attention mechanism.Then the problem of missing modalities is relieved by using the improved loss function to improve the robustness and emotion recognition performance.Experimental results on public datasets show that compared with methods such as Convolutional Neural Network(CNN) and Long Short Term Memory(LSTM) networks,this method has improved the Concordance Correlation Coefficient(CCC) index,and has higher recognition efficiency.It is applicable to dimensional emotion recognition of large volumes of data.
-
Computer Engineering. 2020, 46(6): 73-80. https://doi.org/10.19678/j.issn.1000-3428.0054480The distribution of real complex network nodes obeys power laws,and the hyperbolic geometry can fully represent such characteristics.On this basis,this paper proposes a community division algorithm based on hyperbolic space embedding and minumun clustering.It embeds the modeled complex network into the Poincaré disk model while keeping the global topology information of the complex network.The distribution relationship of nodes is calculated based on the angles on the Poincaré disk to obtain the curve θ.Then the minimum of this curve is selected according to the optimal modularity as the division basis of the optimal community.This paper uses the real access data of China Mobile users to evaluate the effectiveness of the proposed algorithm,and the result shows that,compared with Louvain,SLPA and regularized spectral clustering algorithms,this algorithm does not need to choose a clustering center and its computational complexity is reduced,which has excellent community division performance in real complex networks.
-
Computer Engineering. 2020, 46(6): 81-87. https://doi.org/10.19678/j.issn.1000-3428.0053895To address the high complexity of existing graph representation learning methods,this paper proposes a new graph representation learning method to improve the learning efficiency while maintaining the representation performance of graph features.The method captures the basic properties of graph data by establishing appropriate hyperbolic geometry structure in the neural network representation model.Then the Bayesian Personalized Ranking(BPR) target is used to maximize the gap between the correct links and the wrong links to automatically learn the similarity information.Moreover,the hyperbolic distance function is used to calculate the hierarchical distance between the nodes in the designed neural ranking model.Finally,the model uses the Riemannian gradient descent method to learn the feature vector of nodes.Experimental results show that the proposed method can efficiently learn node features,and can provide more compact and more expressive feature vector representations than DNGR,HARP and other methods.
-
Computer Engineering. 2020, 46(6): 88-93. https://doi.org/10.19678/j.issn.1000-3428.0054038In domain ontology learning,in order to implement optimal clustering of concepts of the same domain without concept overlapping,this paper introduces the graph entropy extreme value theory and proposes a domain concept clustering method.According to the principle of maximum information entropy,the concept nodes of a graph are considered as a whole instead of selecting the centroid.Also,the graph entropy minimization formula is used to design an automatic concept clustering mechanism.Experimental results show that,compared with K-means algorithm,density-based and distance-based domain concept clustering methods,the proposed method significantly improves the precision,recall rate and comprehensive evaluation index,F value.
-
Computer Engineering. 2020, 46(6): 94-102. https://doi.org/10.19678/j.issn.1000-3428.0054210This paper proposes an anomaly community detection method via subspace by combining node attributes with structure information.First,in the given set of to-be-tested communities,the subspace solution strategy based on the average distance of attributes,the subspace inference strategy based on negative entropy weighting and the subspace fusion solution strategy are designed to excavate the attribute weight subspace of each community.Second,the quality assessment model is defined based on the community structure relationships to quantify the community quality scores.Finally,the set of anomaly communities with a low quality score is obtained.Experimental results show that the proposed method can accurately detect anomaly communities,and has better robustness and scalability than AMEN,SODA and other detection methods in artificial network and real network datasets.
-
Computer Engineering. 2020, 46(6): 103-107. https://doi.org/10.19678/j.issn.1000-3428.0054400Siamese network is widely used in the field of target tracking because of its significant advantage in high speed and accuracy.The twin network is composed of two independent branches:semantic branch and appearance branch.Each branch is a twin network with similar learning,which solves the problem of insufficient accuracy of the original twin network.However,each branch is trained independently,which results in the decrease of system speed.To address this problem,this paper proposes ASTS,an adaptive selective system based on twofold siamese network.In the testing process,the network automatically stops propagating forward at the simple frame and rapidly judge the position of the target,so as to improve the tracking speed of the system.In the case of complex frames,the two branches coordinate with each other to track the target accurately.Experimental results on the OTB2013/50/100 and VOT2017 datasets show that compared with the fixed twofold siamese network object tracking method,the ASTS system has faster speed and higher tracking accuracy.
-
Computer Engineering. 2020, 46(6): 108-114. https://doi.org/10.19678/j.issn.1000-3428.0054930Most traditional clustering algorithms need to pre-set clustering parameters and fail to recognize outliers and noise.To address the problem,this paper proposes an adaptive correlation fusion clustering algorithm.The algorithm uses the narual neighbor search algorithm to calculate the density distribution of datasets,and screens out representative kernels with data structure information.The influence of boundary points and noise on clustering results is ruled out.Then the algorithm introduces correlation matrix.By calculating the correlation degree and fusion measurement between clusters,the optimal correlation clusters are selected for fusion to obtain the final clustering result.Experimental results show that compared with Density-Based Spatial Clustering of Applications with Noise(DBSCAN) algorithm and K-means clustering algorithm,the proposed algorithm does not need to manually set clustering parameters,and it has higher clustering accuracy and reliability.
-
Computer Engineering. 2020, 46(6): 115-121. https://doi.org/10.19678/j.issn.1000-3428.0054936Traditional asynchronous acquisition method affects the signal-to-noise ratio of collected power information information,resulting in a decrease in the success rate of power information analysis.To address the problem of asynchronous acquisition,this paper proposes a new power information collection method for clock synchronization.Based on the clock synchronization collection platform of the Field Programmable Gate Array(FPGA),this method uses clock synchronization based devices to send a clock signal of synchronization to the to-be-collected device and the oscilloscope,so that the to-be-collected device is synchronized with the working state of the oscilloscope.The principle of electrical decoupling is used to isolate the influence of external signals on the to-be-collected device,and improves the signal-to-noise ratio of the power consumption information.Through Correlation Power Analysis(CPA),result shows that the proposed method improves the collection efficiency by up to 66.7%,greatly improving the success rate of power analysis.
-
Computer Engineering. 2020, 46(6): 122-129. https://doi.org/10.19678/j.issn.1000-3428.0054050In order to safely transmit information in the physical layer,this paper proposes a physical layer encryption scheme based on constellation obfuscation.The channel coefficient is used as the key and superimposed with the modulated symbol vector to achieve encryption.Considering the actual situation of channel estimation error,this paper analyzes the influence of channel estimation error on the performance of constellation obfuscation encryption scheme and conducts the theoretical formula of bit error rate of receiver with phase estimation error.Simulation results show that the proposed scheme can achieve secure communication and it has certain tolerance for the channel phase error.The system is robust when the channel phase error is with 15°,but when the error exceeds 42°,the system bit error rate is 1.
-
Computer Engineering. 2020, 46(6): 130-135. https://doi.org/10.19678/j.issn.1000-3428.0055219To comprehensively and accurately analyze the security situation of a given network and evaluate the situation,this paper proposes a mixed Network Security Situation Awareness(NASS) model based on Bayesian method.The model preprocesses the situation indicator data collected from a given network environment by discretizing them.Then according to the different evaluation methods,the hierarchical model of situation indicators is established.Finally,the situation influence indicators at the bottom layer of the hierarchical model are merged upward layer by layer by using the Bayesian network model,and the final evaluation index of network security situation is obtained to give the status rating.Experimental results show that the proposed model meets the practical requirements of applications,and the evaluation results are accurate and effective,improving the stability and reliability of network environment.
-
Computer Engineering. 2020, 46(6): 136-143. https://doi.org/10.19678/j.issn.1000-3428.0055554To authorize others to deal with encrypted mails and enable the search of encrypted mails in an encrypted email system,this paper proposes a proxy re-encryption scheme that supports keyword search for email systems.In this scheme,searchable encryption technology is used to search encrypted mails,and then proxy re-encryption technology is used to authorize encrypted mails.Security certification and efficiency analysis results show that the proposed scheme can better resist tampering attacks and keyword offline guessing attacks.At the same time,under the standard model,it is proven that the scheme respectively meets trapdoor privacy security,keyword privacy security and ciphertext privacy security in the determination of the Diffie-Hellman problem,the Decisional Bilinear Diffie-Hellman(DBDH) problem,and the Quotient Decisional Bilinear Diffie-Hellman(QDBDH) problem.Compared with the dPRES scheme,the proposed scheme reduces the time cost and improves the efficiency of search and decryption.
-
Computer Engineering. 2020, 46(6): 144-148. https://doi.org/10.19678/j.issn.1000-3428.0054421The authentication encryption algorithm is difficult to generate the indistinguishability of Chosen Plaintext Attack(CPA) in practical application.Therefore,an improved encryption scheme of adjustable authentication is proposed by means of hardware algorithm platform,sign sequence number of application data packet and dynamic adjustable key counter.The algorithm can run smoothly on the network by adding small hardware components in exchange for the parallel computation of adjustable factor and encryption,which supports the application of resource-limited embedded devices.Experimental results show that this method can shorten the whole runtime of the system and improve the overall running efficiency.
-
Computer Engineering. 2020, 46(6): 149-154. https://doi.org/10.19678/j.issn.1000-3428.0055014Licensed Auxiliary Access(LAA) is a new carrier aggregation operation mode for Long-Term Evolution(LTE) advanced systems to transmit data over authorized and unauthorized spectrums.To coexist with WiFi systems,LAA need to perform pre-call listening,which significantly affects their resource utilization.To address the problem,this paper proposes a Multi-Subframe Scheduling(MSS) to improve the performance of uplink LAA data transmission.MSS allows user equipment to have multiple channel sensing and measurement opportunities to continuously transmit data of multiple subframes.The best parameter configuration of MSS that maximizes the utilization of LAA resources is provided,and the MSS scheme is compared with scheduling-based schemes.Results show that the proposed MSS scheme can significantly improve the resource utilization of unlicensed spectrum.
-
Computer Engineering. 2020, 46(6): 155-163. https://doi.org/10.19678/j.issn.1000-3428.0054992In order to realize efficient and reliable network transmission protocol,according to RUDP draft,this paper introduces the reliable TCP mechanism to design a high speed and reliable network transmission protocol ORUDP based on message packet for connection.The basic protocol is implemented by establishing the confirmation mechanism,retransmission mechanism,flow control mechanism and double queue acceleration mechanism.Then Field Programmable Gate Array(FPGA) is selected for the logical design and implementation of the ORUDP protocol stack,and the functional simulation of all design modules is completed by using the simulation tool,Modlesim.A test environment is built to test the ORUDP network protocol stack,and the result shows that ORUDP can reliably transmit data packets,solving loss of data packets,out-of-order data packets,and repeated arrival of data packets.It keeps a high transmission speed while fewer resources are consumed and the content of packets is short.
-
Computer Engineering. 2020, 46(6): 164-171. https://doi.org/10.19678/j.issn.1000-3428.0054767Millimeter-wave small cell networks can provide high regional throughput,but its limited return link capacity often results in block.To address the problem,a content delivery method based on the combination of cache Small Base Station(SBS) and wireless self-backhaul SBS is proposed.Some popular content is pre-stored in the cache,and the content that is not stored is transmitted by the Macro Base Station(MBS) with deployed Multi-Input and Multi-Output(MIMO).Then the expressions of coverage probability,Average Area Rate(AAR) and average area Energy Efficiency(EE) of heterogeneous network of cache and backhaul joint content delivery are derived by using the millimeter wave antenna mode,the horizon sphere model and the stochastic geometry tool.Experimental results show that this method makes the beam narrower and the transmission more directional,and improves the performance of average area rate and access link of self-backhaul transmission.
-
Computer Engineering. 2020, 46(6): 172-177. https://doi.org/10.19678/j.issn.1000-3428.0055094To address increased node energy consumption and decreased data accuracy in Wireless Sensor Network(WSN),this paper proposes a Data Fusion Algorithm Based on Game Theory(DFABGT).DFABGT adopts a clustered network structure.The nodes in the cluster decide the benefit function based on the game between benefit and energy consumption,so as to select the nodes with the lowest energy consumption.Then maximum value of the benefit function serves as the weight in the calculation of confidence distance to obtain reliable data.Then the head node of the cluster transmits the reliable data collected by the nodes in the cluster to the Sink node,and the Sink node uses the Bayesian theory to complete data fusion. Experimental results show that compared with the E-CPDA algorithm,the MGDAA algorithm and the Megrez algorithm,the proposed algorithm increases the fusion data accuracy by 3.9%,21.2% and 12.1% respectively,and reduces the energy consumption of nodes by 28%,22% and 19% respectively.
-
Computer Engineering. 2020, 46(6): 178-186. https://doi.org/10.19678/j.issn.1000-3428.0054643To address network coverage holes caused by uneven distribution of randomly deployed sensor nodes,this paper proposes a dynamic detection and repair algorithm for three-dimensional coverage holes.In the target monitoring area where the hybrid nodes are randomly deployed,the three-dimensional space is divided into cube mesh,and the coverage holes are detected according to the edge endpoints and edge arcs of the selected nodes.Then for the redundant mobile nodes around the coverage holes,their distance from the coverage holes and the moving direction are calculated,and then the nodes are adjusted to repair the coverage holes.Experimental results show that the proposed algorithm has higher utilization of nodes,lower network coverage costs,and needs fewer nodes to meet overall network coverage requirements compared with PSO,CPA and other algorithms.Also,the algorithm reduces the energy consumption in movement.
-
Computer Engineering. 2020, 46(6): 187-195. https://doi.org/10.19678/j.issn.1000-3428.0054723In order to balance and save the energy consumption of nodes in Wireless Sensor Network(WSN),this paper proposes a clustering routing algorithm based on hierarchical neighboring nodes.Considering the status of neighboring cluster heads and neighboring nodes,it performs hierarchical processing on the nodes,and re-optimizes the cluster heads according to their position and the range of the cluster group in the cluster selective stage to avoid over-dense distribution of cluster heads and unreasonable cluster group range.In the stage of data transmission,through the remaining energy of the nodes,the distance between nodes and neighboring nodes,the relay mode is used to balance the energy consumption of the nodes and reduce the total energy consumption of network.Simulation results show that compared with LEACH,DEEC and CECA lamp algorithm boxes,the proposed algorithm can better reduce and balance network energy consumption and extend network lifetime.
-
Computer Engineering. 2020, 46(6): 196-201. https://doi.org/10.19678/j.issn.1000-3428.0056220In millimeter-Wave(mmWave) communication network of Unmanned Aerial Vehicles(UAV),communication between the UAV and the base station provides a critical backhaul link for interactions between the UAV and the core network.In order to support stable communication with a high transmissionrate between UAV and base stations as well as fast access,this paper proposes a position-aided channel estimation method for mmWave communication of UAV.With the goal of improving the performance of mmWave network serving UAV,the three-dimensional position model of UAV and mmWave base stations is constructed.Meanwhile,hierarchical multi-resolution codebook and adaptive channel estimation method are adopted,and the side information from the global satellite navigation system assists in channel estimation.Then performance analysis is implemented on the received Signal-to-Noise Ratio(SNR) and the average channel estimation time.Simulation results show that,compared with the methods without the assistance of position information,the proposed method can effectively accelerate the channel estimation process and obtain higher antenna array gains.
-
Computer Engineering. 2020, 46(6): 202-208. https://doi.org/10.19678/j.issn.1000-3428.0055016To address the low operation efficiency and poor operability of traditional defect location methods of software testing in large software systems,this paper proposes a defect location method based on risk trace and complex network.First,the sequence of function calls of a program is obtained dynamically.Next,based on the execution results of test cases on programs with different versions of defects,the target sequence and suspicious sequence of the to-be-tested program are selected for comparison,so as to find the risk trace and extract the set of suspicious functions.Then the suspicious function set is checked.If no defect function is found,a complex network graph is built for the to-be-tested program and sorted based on the out-degree of function nodes to rule out functions that have been checked.Finally,a set of candidate defect functions is generated,and the defect function is located.Experimental results show that compared with the existing Combine and Upper methods,the proposed method canimprove the efficiency of defect location by 22.2% and 12.5% respectively,and it is more operable on large software systems.
-
Computer Engineering. 2020, 46(6): 209-215. https://doi.org/10.19678/j.issn.1000-3428.0054783In order to enable the Software Defined Network(SDN) to cope with the possible situations of single link failure in network,this paper proposes a routing protection algorithm based on hybrid SDN.The routing protection algorithm coping with the single link failure is deployed in the hybrid SDN,which is reduced to a 0-1 integer programming problem and a heuristic algorithm is used to calculate the approximate optimal solution to the problem.The execution process of the algorithm is introduced through an example and the corresponding time complexity of the algorithm is analyzed.Experimental results show that the proposed algorithm can cope with all possible situations of single link failure only by upgrading a small number of nodes in the traditional network to SDN nodes,and the corresponding path stretch is within 1.4.
-
Computer Engineering. 2020, 46(6): 216-220. https://doi.org/10.19678/j.issn.1000-3428.0054974Testing based on operational profile is designed to find failures of high incidence rate,but it is not suitable for systems that has rule out failures of high incidence rates and requires high reliability.To address the problem,this paper proposes an allocation and selection method of test cases based on operational profile by improving traditional operational profile testing.The method uses the adaptive importance sampling method to dynamically change the probability of test cases in each iteration,and automatically makes adjustments according to the test results to select more efficient test cases to improve the effectiveness of the test and the reliability of the software.Analysis results of test cases in five subdomains verify the effectiveness of the proposed method.
-
Computer Engineering. 2020, 46(6): 221-229. https://doi.org/10.19678/j.issn.1000-3428.0054565To address the scalability problem of control plane in software defined aeronautical information network,this paper proposes a multi-controller deployment scheme.The scheme contains two phases,cluster domain partitioning and intra-domain controller deployment.In the cluster domain partitioning phase,to deal with the unstable initial convergence of the k-means algorithm,an improved k-means algorithm based on Discrete Factor(DF) is designed to partition the aeronautical information network into multiple aeronautical cluster domains.In the intra-domain deployment phase,a discrete particle swarm optimization algorithm is adopted for controller deployment with the goal of minimizing the failure rate of network control paths,so as to implement effective management and control of network.Simulation results show that the proposed scheme can achieve reasonable partitioning of aeronautical information network with load balancing of controllers ensured.The adopted discrete particle swarm optimization algorithm can effectively reduce the failure rate of control paths,and solve the multi-controller deployment problem in dynamic and large-scale networks.
-
Computer Engineering. 2020, 46(6): 230-240. https://doi.org/10.19678/j.issn.1000-3428.0054070The ocean computing model has the characteristics of wide data sources and high model complexity,and involves many disciplines.In order to solve the complex problem in the collaborative computing of multiple ocean computing models,a collaborative integrated process management system for ocean computing models is developed.The model is integrated as services,an universal data conversion interface is designed,a data classification and conversion algorithm based on Random Forest(RF) is proposed to realize the collaborative conversion of ocean data,and the time complexity is reduced by adding a data preprocessing process.On this basis,an improved Chicken Swarm Optimization(CSO) algorithm is designed to improve scheduling efficiency,and a lightweight process customization interaction mechanism is constructed by using the service-oriented multi-granularity collaborative process modeling method.Experimental results show that the system can effectively improve the efficiency of the ocean data analysis and numerical simulation.The RF algorithm combining with data preprocessing has faster data classification speed compared with SVM and the original RF algorithm,and can keep the classification accuracy higher than 91%.Compared with the original CSO algorithm and SJC algorithm,the iterations of the improved CSO algorithm is reduced by 29%~37%,which can effectively improve the scheduling efficiency.
-
Computer Engineering. 2020, 46(6): 241-247. https://doi.org/10.19678/j.issn.1000-3428.0054809To improve the accuracy and contrast of the saliency graph generated by saliency detection models,this paper proposes a saliency detection method based on relation between boundary and center.First,the image is smoothed by guided filters and segmented by SLIC superpixel.Second,the saliency of the superpixel block is calculated according to the relation between the center point and the boundary point,and salient figure 1 is obtained by using background suppression through gamma transformation.At the same time,according to the relationship between the boundary point and the center point,the seed points are obtained,and the manifold ranking algorithm is improved.On this basis the salient figure 2 is obtained by using background suppression through gamma transformation.Finally,the two salient images are fused at the pixel level to obtain the final salient image.Experimental results show that the proposed method outperforms existing methods like COV,DSR and GR in terms of F-Measure,E-Measure,MAE and other indicators,improving background suppression effects.
-
Computer Engineering. 2020, 46(6): 248-255. https://doi.org/10.19678/j.issn.1000-3428.0056417In order to improve the accuracy of the target detection model using convolutional neural network and enhance the detection ability of the detector for small targets,this paper proposes a multi-scale target detection network model trained from scratch.The detection network is trained from scratch to increase its accuracy to the level of pre-trained models or even higher.Then a new Deformable-ScratchNet network model is designed according to the characteristics of small targets.Its network structure is adjusted,and shallow information is integrated with the model to improve the detection performance of small targets.Experimental results show that compared with Faster-RCNN and other classic network models,the proposed model has higher detection accuracy on the PASCAL VOC data set and self-made remote sensing image of military target data set.
-
Computer Engineering. 2020, 46(6): 256-265. https://doi.org/10.19678/j.issn.1000-3428.0055351To solve the problem of lack of image description of spatial relationship in construction scenes,this paper proposes an image description generation model integrating construction scenes and spatial relationship.YOLOv3 network is used for target detection,and a feature extraction layer is added to the traditional object detection model on the basis of the transE algorithm to form the relationship detection model.The coordinate frame information of the object is combined to obtain the relationship between the objects,and the method based on rules and templates is used to generate the image description.Experimental results show that compared with m-RNN,NIC,Soft-Attention and Hard-Attention models,the proposed model can generate accurate description of spatial relationship.
-
Computer Engineering. 2020, 46(6): 266-273. https://doi.org/10.19678/j.issn.1000-3428.0056188In terms of medical treatment,the diagnosis of many diseases relies on the observation of microscopic objects such as cells with a high magnification microscope.However,due to the high price and complex operation of high magnification microscope and there are some problems in the reconstruction of high magnification cell micro-images,such as the inconsistency of image style between high magnification micro-images and low magnification micro-images,the different resolution of cell images and the lacking of paired training data.To solve the above problems,a high magnification cell micro-images generative adversarial network is proposed.Based on the CycleGAN,a new residual dense block is added to the generator while the new activation function is introduced,and the Batch Normalization(BN) layers are removed.At the same time,in order to ensure the authenticity of the generated images,the detail perceptual loss is introduced to the training process of the generator.Experimental results show that the proposed method can effectively restore the detail of the high magnification micro-images while preserving the basic information of the low magnification micro-images.
-
Computer Engineering. 2020, 46(6): 274-280. https://doi.org/10.19678/j.issn.1000-3428.0054566Face detection is an important research direction in computer vision and information security,which has been widely studied over the past few decades.In the traditional face detection method,there is no pixel-level segmentation process,which leads to the problem of face features with noise and unsatisfactory detection accuracy.In order to overcome this shortcoming,a face detection and segmentation method based on Mask R-CNN is proposed in this paper.In this method,ResNet-101 and RPN is used to generate RoIs,and RoIAlign faithfully retains the exact spatial locations to generate binary mask through Fully Convolution Network.In order to train the model,this paper constructs a face dataset with segmentation annotation information.The experimental results of well-known face detection dataset show that the proposed method has better face detection effect and can achieve pixel-level face information segmentation at the same time.
-
Computer Engineering. 2020, 46(6): 281-287. https://doi.org/10.19678/j.issn.1000-3428.0054730Aiming at the identification problems that Smeared Spectrum(SMSP),Chopping and Interleaving(C&I),smart noise jamming,the composite jamming of noise amplitude modulation and noise range deception,and the composite jamming of noise frequency modulation and noise range deception.This paper proposes a SAE-GA-SVM-based identification model algorithm,which can identify.The algorithm constructs a mathematical model for target echo and jamming signals,employing a multi-domain joint feature extraction method to extract the 47-dimensional features.In order to effectively remove redundant data and maintain a high identification rate,the Stacked Auto-Encoder(SAE) algorithm in deep learning is adopted.By using the SAE structure,a mutual mapping between high-dimensional space and low-dimensional space is constructed to obtain the corresponding optimal low-dimensional representation of raw data.Then the Genetic Algorithm(GA) is used to optimize the penalty factor and kernel function parameters of Support Vector Machine(SVM),and on this basis the SAE-GA-SVM-based model for new radar jamming identification is established.Simulation results show that the proposed model can effectively reduce the feature dimension,and its classification accuracy is 10% higher than that of the traditional detection models.
-
Computer Engineering. 2020, 46(6): 288-295. https://doi.org/10.19678/j.issn.1000-3428.0055232The social risk caused by prevalence of zombie fans brings significant threat to the credibility of social platforms.To effectively recognize these zombie fans,this paper proposes a zombie fans recognition model,Zat-NN,based on neural network.First,the social behavior of zombie fans on microblog is analyzed to obtain behavior features of high-level zombie fans.Second,the cumulative distribution function is used to study the behavior feature differences between zombie fans and normal users.Then Convolutional Neural Network(CNN) and Long Short Term Memory(LSTM) network are combined to strengthen the sentiment analysis of microblog texts.At the same time,the number of daily forwarded microblogs,blogging tools and microblog emotion features are added as user features to improve the recognition accuracy and robustness of the Zat-NN model.Experimental results on the user dataset of Sina microblog show that the Zat-NN model can effectively recognize high-level zombie fans,improving user experience of social network.
-
Computer Engineering. 2020, 46(6): 296-302. https://doi.org/10.19678/j.issn.1000-3428.0054431Most of existing methods for entity relationship recognition take a single sentence as processing unit,and fail to address tagging errors of entity relationships in the training corpus.Also,they cannot make full use of the mutual reinforcement of multiple sentences that contain entity information in relationship recognition.Therefore,this paper proposes a recognition method based on bidirectional Gated Recurrent Unit(GRU) and dual attention mechanism for entity relationships of Chinese electronic medical records.This paper proposes a BiGRU-Dual Attention model,and uses bidirectional GRU to learn the context information of characters in order to obtain more fine-grained features.Then the character-level attention mechanism is introduced to improve the weight of the characters that are key to relation recognition.Also,the sentence-level attention mechanism is employed to capture the features that can enhance recognition performance from multiple sentences,so as to reduce the weight of mislabeled sentences.Experimental results show that compared with the mainstream BiLSTM-Attention model,the proposed model increases the F1 value by 3.97% to 82.17%.
-
Computer Engineering. 2020, 46(6): 303-307. https://doi.org/10.19678/j.issn.1000-3428.0055114The traditional negative selection process takes a long time to match all detectors with test data to eliminate abnormal data,resulting in low detection efficiency.Therefore,this paper proposes a negative selection algorithm based on hierarchical clustering of the detector set.The number of detectors that need to calculate the distance is reduced by hierarchical clustering of the generated detectors.The data that does not match the detector is no longer directly marked as normal data,but is marked based on the calculation results of the distance between the data and the self-set and the detector set.Experimental results show that compared with the V-detector algorithm and the real-valued negative selection algorithm of immunity,the proposed algorithm significantly improves the detection efficiency and reduces the false detection rate.
-
Computer Engineering. 2020, 46(6): 308-313. https://doi.org/10.19678/j.issn.1000-3428.0056250In order to improve the adaptability to abnormal load judgment in scenarios where loads fluctuate greatly,this paper proposes a threshold model with convolutional neural network suitable for the dynamic judgment of abnormal load of power grid.The Convolutional Neural Network(CNN) model is trained with historical load data in time series for load prediction,and based on the predicted loads,the future state variable data in power grid can be calculated.Based on the construction of the state variable data source matrix,its window matrix,standard matrix and sample covariance matrix are subsequently obtained.Then the dynamic threshold based on the maximum eigenvalue of the sample covariance matrix is set,and the threshold is used for the overdue judgment of the maximum eigenvalue at the current time,so as to implement dynamic judgment of abnormal load in power grid.With the help of software tools including Matlab R2014a and PST,simulation tests are performed on the IEEE50 machine 145 bus standard system.Results show that compared with the traditional threshold model,the proposed threshold model is more adaptable and accurate for the judgment of abnormal MESCM indicators in dynamic power grid.
-
Computer Engineering. 2020, 46(6): 314-320. https://doi.org/10.19678/j.issn.1000-3428.0055420In order to meet the requirements of edge devices for accurate,fast and adaptive generated parameters of machine learning in the Internet of Things(IoT),this paper proposes an online GP-ELM algorithm based on the differential-evolution extreme learning machine,this paper proposes an online GP-ELM algorithm.The algorithm improves the way of adding nodes by carrying on node statistics and deleting nodes while everytime a node is added,so as to increase the training speed,while maintaining the accuracy of the algorithm.Matlab software is used to train the dataset image segmentation,satellite image classification,satellite DNA and conduct experiments.Results show that compared with EI-ELM,D-ELM,EM-ELM and other algorithms,GP-ELM algorithm has better performance in accuracy,training time,model size and generalization ability.
2024
Volume 50 No 3
Volume 50 No 3
(Started in 1975, Monthly)
Sponsored by：
East-China Institute of Computer Technology Shanghai Computer Society
Edited & Published by：
Editorial Board of Computer Engineering
Chief Editor：Jiang Bo
Executive Chief Editor:
Hu Bing
Tel：021-67092217
E-mail: ecice06@ecict.com.cn
CN: 31-1289/TP
ISSN: 1000-3428
Sponsored by：
East-China Institute of Computer Technology Shanghai Computer Society
Edited & Published by：
Editorial Board of Computer Engineering
Chief Editor：Jiang Bo
Executive Chief Editor:
Hu Bing
Tel：021-67092217
E-mail: ecice06@ecict.com.cn
CN: 31-1289/TP
ISSN: 1000-3428