15 August 2018, Volume 44 Issue 8
-
Select all|
-
Computer Engineering. 2018, 44(8): 1-6. https://doi.org/10.19678/j.issn.1000-3428.0048571
The traditional Ciphertextpolicy Attribute-based Encryption(CP-ABE) systems are based on singleauthority organization,and most of them have the problems of large computation cost,low key maintenance efficiency,and inability to achieve anti-conspiracy attacks.Therefore,an improved cloud access control scheme is proposed.The logical binary tree constructed by the hash function generates a group key for each attribute,and uses the group key to update the user’s private key and ciphertext to achieve fine-grained attribute changes.The private key of the user’s attribute is jointly distributed by multiple authorized centers to solve the performance bottleneck of the single authorized organization.With decryption outsourcing and fixed ciphertext encryption technology,users’ computing time and storage cost are reduced.The introduction of linear secret sharing matrix achieves a flexible resource access control strategy.Analysis results show that compared with other schemes such as cloud storage schemes based on attribute encryption,the complexity of the user’s computation is optimal when the attribute changes,which greatly enhances the user’s decryption efficiency.
-
Computer Engineering. 2018, 44(8): 7-13. https://doi.org/10.19678/j.issn.1000-3428.0048415
For cloud data access control research,there are many problems such as large cost of attribute change and inflexibility.On the basis of ciphertext policy Attribute-based Encryption(ABE),this paper proposes a scheme supporting fine-grained attribute change.It combines a Hash function that has irreversible calculation properties and defines a logical binary tree to re-encrypt the ciphertext,and updates the ciphertext and the private key according to the group key generated by the root node of the largest covering subtree corresponding to each attribute,so that each attribute can be fine-grained changed.Simulation experimental results show that this scheme can effectively reduce the whole computation complexity and user storage pressure,and the change efficiency can be improved.
-
Computer Engineering. 2018, 44(8): 14-18. https://doi.org/10.19678/j.issn.1000-3428.0049169
Aiming at the problem that task scheduling based on ant colony algorithm has unbalanced load and slow convergence speed,an improved task scheduling optimization algorithm is proposed.The pheromone update rules of the ant colony algorithm are optimized by weighting methods to accelerate the solution speed,and the comprehensive performance of the dynamic update volatilization coefficient optimization algorithm is utilized,and the load weight coefficient of the virtual machine is introduced during the update process of the local pheromone to ensure the load balancing of virtual machines.Experimental results show that the task scheduling strategy of the improved algorithm ensures that the task is reasonably allocated,and at the same time,the convergence speed of the algorithm is improved and the total execution time is shortened.
-
Computer Engineering. 2018, 44(8): 19-23,29. https://doi.org/10.19678/j.issn.1000-3428.0049268
For the attribute-based encryption scheme,a single authority has the problem of low efficiency and easy disclosure of user keys.Through the establishment of a decentralized attribute authority and agent reencryption technology,a Ciphertext-policy Attribute-based Encrypfion(CP-ABE) with attribute revocation function in a multi-attribute authority environment is designed.The scheme implements user attribute revocation through the version number tagging method,and establishes the access tree structure to manage attributes issued by different attribute authorities,so as to resist the joint attack of multiple attribute authority and achieve fine-grained access of users.Analysis results show that compared with the existing attribute-based encryption schemes,this scheme can shorten the length of the ciphertext and reduce the user’s computation cost.
-
Computer Engineering. 2018, 44(8): 24-29. https://doi.org/10.19678/j.issn.1000-3428.0047251
In data decryption process of cloud platform server,it is easy to cause problems of information disclosure,the database administrators spy on users’ privacy information and heavy processing tasks of client.To solve above problems,a design scheme of security private cloud platform based on homomorphic cryptography is proposed.In the private cloud platform scene meeting the demand of bank business processing,by adopting with adjustable onion encryption strategy,ciphertexts can be evaluated directly by additive homomorphism of Paillier and multiplicative homomorphism of ElGamal.Then,frequent interactions between client and server,as well as data encryption and decryption operations can be avoided.Simulation experiment and efficiency analysis results show that the proposed scheme can effectively resist the chosen plaintext attack while ensuring the operation efficiency.
-
Computer Engineering. 2018, 44(8): 30-37. https://doi.org/10.19678/j.issn.1000-3428.0047754
To optimize the scheduling process of cloud workflow tasks,a workflow scheduling algorithm based on deadline constraint and Critical Path(CP) is presented,which named WS-DCCP.Firstly,combined with the heterogenous and elastic features of cloud resources,all tasks are partitioned into different logic levels.Based on this logic levels,the workflow deadline is proportionally re-distributed.Then,tasks are prioritized by the improved sum of task upwark rank and downward rank,and the Constrainted CP(CCP) are constructed based on task priority.Finally,the tasks set on the CCP are scheduled on the same resource so as to reduce the communication cost.The next is to find the resource minimizing the workflow execution cost under meeting the sub-deadline constraint of the CCP,which can obtain the scheduling solution minimizing the execution cost under meeting deadline constraint.Through the simulation experiments,the performance evaluations are conducted compared with IC-PCP and JIT.The results show that WS-DCCP can reduce the workflow scheduling cost and improve the scheduling success rate while meeting deadline constraints.
-
Computer Engineering. 2018, 44(8): 38-42,47. https://doi.org/10.19678/j.issn.1000-3428.0049514
For the classification problem of the triangular fuzzy number,this paper constructs similarity-based Fisher linear discriminant analysis model.The model is the generalization of the classical Fisher linear discriminant analysis model.The similarity of triangular fuzzy number is employed for discrete degree within class,with a distance of triangular fuzzy number to quantify between each class of dispersion degree.At the same time,an idea consists of finding a projection direction vector which maximizes the radio of the dispersion degree between the class data and within the class is proposed.The results of instance analysis show that the proposed classification model is effective and applicable in the diagnosis of gastric diseases.
-
Computer Engineering. 2018, 44(8): 43-47. https://doi.org/10.19678/j.issn.1000-3428.0048099
A new list scheduling algorithm called IFEFT is proposed based on HEFT algorithm for minimizing makespan which can be modelled as a Directed Acyclic Graph(DAG) for static task scheduling problems in heterogeneous distributed computing environment.In the phase of processor selection,a better processor selected by changing the policy of HEFT processor is allocated based on the product of earliest finish time and link distance which makes value smaller to assign the processor for considering the impact of the task on its immediate predecessor and immediate successor tasks finish time.Finally,based on randomly generated graphs,the comparison studies demonstrate that IFEFT algorithm has an advantage over the HEFT,DLS and CPOP in terms of scheduling efficiency.
-
Computer Engineering. 2018, 44(8): 48-53,60. https://doi.org/10.19678/j.issn.1000-3428.0047852
In Distributed Shared Memory(DSM),using multicast to transmit messages between nodes can greatly save communication costs,but traditional multicast is uncontrollable.To solve this problem,a scheme to apply multicast based on Software Defined Network(SDN) to DSM is proposed.A DSM system based on multicast is designed,and data transmission is provided for it with SDN based multicast.Experimental results show that,compared with traditional Ethernet,this method can effectively reduce the data transmission delay in DSM and provide good communication performance guarantee for DSM system.
-
Computer Engineering. 2018, 44(8): 54-60. https://doi.org/10.19678/j.issn.1000-3428.0048047
In order to solve large-scale sparse group of linear equations based on Graphic Processing Unit(GPU),in this paper,a storage format HEC(Hybrid ELL and CSR) of sparse matrix is proposed to solve sparse linear equations on GPU.This format is successfully used in Compute Unified Device Architecture(CUDA) to realize Incomplete LU factorization preconditioned Conjugate Gradient(ILUCG)method.It consists of ELL and CSR(Compressed Sparse Row) and is applied to solve the large sparse linear systems by calling GPU kernel in ILUCG.The storage efficiency of sparse matrices can be improved and operation time of Sparse Matrix-Vector multiplication(SpMV) can be reduced.The way by calling GPU kernel and being stored in HEC is compared with which is carried out by calling CUSPARSE library functions based on CSR and HYB(Hybrid).The result shows that the acceleration of the best available is 10.4%,and the way by using HEC storage format has good SpMV performance.
-
Computer Engineering. 2018, 44(8): 61-67,73. https://doi.org/10.19678/j.issn.1000-3428.0050805
The current mobile target behavioral rules mostly focus on the single dimension of space or time,or the self-behavior of a single moving target.In order to find out the association co-occurrence rule of moving targets,based on the traditional trajectory sequential pattern mining,this paper adds task attribute information with commonly used time stamps and place association models,and then takes the co-occurrence pattern of moving objects in space as a benchmark to further analyze the frequency of time,proposes a pattern mining algorithm which applies the association rule and sequence mining to the location and equipment association data of the moving target.Experimental results prove that this algorithm can effectively mine the activity location correlation and the rule of co-occurrence of moving target,and has lower algorithm complexity,which can provide decision-making for the establishment of the reaction mechanism of the moving target.
-
Computer Engineering. 2018, 44(8): 68-73. https://doi.org/10.19678/j.issn.1000-3428.0049691
At present,in the research field of social network trust degree prediction,most proposed models or algorithms just are suitable for small-scale network,or have bad efficiency in large-scale network.To solve this problem,considering the node topological structure information and the user trust rate information in the network,a trust degree prediction algorithm for large-scale social network is proposed.The large-scale social network is reduced to a trust graph,and the trust degree between users is calculated on the trust graph.The topological structure information of nodes can be calculated under line.Experimental results show that,compared with the typical trust prediction algorithm Tidal Trust and SWTrust,the proposed algorithm has higher prediction accuracy and computational efficiency.
-
Computer Engineering. 2018, 44(8): 74-78. https://doi.org/10.19678/j.issn.1000-3428.0047421
When using the shortest path control rule to design reliability of control network,it is easy to choose high failure rate link,resulting in unreliable control network.Therefore,taking link reliability index as a control rule,a scheme of improving the reliability of control network is proposed.By using the controller location deployment algorithm,we choose a reasonable location in the network to deploy the controller,and the controller reaches the OpenFlow switch with the most reliable link.Simulation results show that,in terms of reliability index,the scheme improves the management mode of commonly used control network up to 85%.
-
Computer Engineering. 2018, 44(8): 79-85. https://doi.org/10.19678/j.issn.1000-3428.0048168
The traditional indoor personnel behavior detection method has low accuracy and poor stability.To solve this problem,an indoor personnel behavior detection method based on Channel State Information(CSI) is proposed.After collecting the CSI raw data package,it uses Kalman filtering algorithm to filter it,and uses the SVM algorithm to classify the filtered data,and then builds the fingerprint database.At the same time,the PSO algorithm is used to modify the parameters of the SVM,then the real time data collected from the real environment is processed by SVM,and the real-time data is matched with the data of the fingerprint library.On this basis,indoor personnel behavior detection is realized.Experimental results show that,compared with LIFS and FIMD method,this method can identify the behavior of indoor personnel more accurately and has higher stability.
-
Computer Engineering. 2018, 44(8): 86-92,99. https://doi.org/10.19678/j.issn.1000-3428.0048269
Aiming at the problem of high energy consumption in wireless communication,based on multiuser cognitive Multiple-Input Multiple-Output(MIMO) networks system and using Time Division Multiplexing-Code Division Multiple Access(TDM-CDMA) technology,the "best time & multiuser sharing" method is proposed to allocate time resources to sub user groups in parallel.Firstly,the mathematical model of multiuser cognitive MIMO communication system with parallel time allocation is established by using TDM-CDMA method,and the model is nonconvex problem.Then,the nonconvex problem is transformed into the time-related convex optimization problem by water-filling algorithm and SDP relaxation.Secondly,the greedy algorithm is used to solve the convex optimization problem,on this basis,the minimum initialization distribution and the best time allocation of time resources are carried out.Simulation results show that compared with the traditional TDMA technology to allocate time resources,this method can reduce the energy consumption in the cognitive MIMO system networks.
-
Computer Engineering. 2018, 44(8): 93-99. https://doi.org/10.19678/j.issn.1000-3428.0046771
Sensor nodes in wireless sensor network(WSN) are prone to failure due to energy depletion and the effects of harsh environments.The failure of the node will make the network to be disconnected.This paper studies connectivity restoration of WSN,analyzes and summarizes the related main direction and the research results in recent years.According to the scale of network failure,the restoration algorithms are divided into two types,one is for small-scale failure and the other is for large-scale failure.In addition,the repair algorithms are further classified from other aspects,such as triggering conditions,realization viewpoint,and etc.Finally,the shortcomings of the current research and future research directions are pointed out.
-
Computer Engineering. 2018, 44(8): 100-104,111. https://doi.org/10.19678/j.issn.1000-3428.0048129
In the Train Control and Service Network(TCSN),in order to ensure the priority of the transmission of control services and ensure the quality of service of other control and service business,the traffic flow of Assured Forward(AF) must be monitored.Traditional traffic monitoring methods are static,and can not adapt to changes in network load status.To solve this problem,a dynamic token allocation algorithm for TCSN multi-level token bucket traffic is proposed.On the basis of the traditional token bucket traffic supervision,the two level token allocation algorithm is used to dynamically adjust the token adding rate and redistribute the token bucket,so as to realize the dynamic regulation of the AF control and service traffic flow.Simulation results show that this algorithm can dynamically adjust the token allocation quota according to the number of missing tokens,delay and packet loss rate of AF subqueue under different network conditions.
-
Computer Engineering. 2018, 44(8): 105-111. https://doi.org/10.19678/j.issn.1000-3428.0046523
In Wireless Sensor Network(WSN),in order to locate the nodes accurately,it is needed to deploy anchor nodes reasonably.The influence on positioning accuracy of geometry formed by anchor nodes and positioned nodes is derived in theory.This paper proposes the concept of average area localization error bound as the evaluation criterion of placement performance.The effect of deployment centroid,geometry and geometric area on area location accuracy of the anchor node is analyzed.Based on this,a heuristic algorithm for optimal deployment of static anchor nodes is given.Simulation results show that,compared with classical anchor node search deployment algorithm,the localization accuracy is similar while the computation cost of the proposed algorithm is reduced significantly.
-
Computer Engineering. 2018, 44(8): 112-119,126. https://doi.org/10.19678/j.issn.1000-3428.0050266
In order to improve the packet processing performance of Virtual Network Function(VNF),this paper comes up with a Field Programmable Gate Array(FPGA)-based General Hardware Accelerator(GHA) architecture.The GHA architecture implements the packet processing pipeline on a dynamically reconfigurable FPGA,increasing VNF throughput and ensuring independence between different VNF accelerators.The Discrete Particle Swarm Optimization(DPSO) algorithm is utilized to optimize the allocation of acceleration resources.Experimental results indicate that GHA architecture can increase the throughput of VNF by 50.7 times approximately,and the optimization rate of the acceleration resource allocation strategy can reach 24.5%.
-
Computer Engineering. 2018, 44(8): 120-126. https://doi.org/10.19678/j.issn.1000-3428.0048114
In order to guard the communication quality of mobile users and simultaneously improve the coverage performance of network,a cell split scheme for Heterogeneous Network(HetNet) is proposed,the Macrocell User (MU) is split into Cell Center User(CCU) and Cell Edge User(CEU).The available spectral bands are divided into Cell Center Region(CCR) band and Cell Edge Region(CER) band.The CCR band is shared by CCU and Femtocell User(FU),and the CER band is shared by Device-to-Device(D2D) users and CEU.By modeling the network elements as independent Poisson point process and using stochastic geometry method,the classification of MU and random spectrum access policy are introduced.Experimental results show that using the presented cell split strategy cannot only decrease the user interference but also enhance the coverage performance of CCU and CEU.The proposed scheme can guide for the design of HetNet.
-
Computer Engineering. 2018, 44(8): 127-134. https://doi.org/10.19678/j.issn.1000-3428.0050287
Due to the characteristics of random,flexible,and complex arrangement of nodes in multi-tier heterogeneous cellular network,intra-tier and inter-tier interference cannot be ignored,which becomes more complicated when analyzing the uplink network performance.So this paper considers the structure characteristic and the practical multi-channel environment of the multi-tier heterogeneous network,combines cell bias and frequency division scheme,based on stochastic geometry theory,the user of multi-tier heterogeneous cellular network uplink average ergodic rate is derived in multi-channel environment.Experimental results show that using the frequency division scheme can improve the average upstream ergodic speed of users in multi-tier heterogeneous network.
-
Computer Engineering. 2018, 44(8): 135-141. https://doi.org/10.19678/j.issn.1000-3428.0050521
In order to solve the problem of poor control path reliability and unbalanced load of controllers in the existing multi-controllers placement scheme,a Multi-Controllers Elastic Placement(MCEP) algorithm based on reliability and load optimization is proposed.Based on spectral clustering,the problem of multi-controllers placement is transformed into a row vector classification problem of network topology matrix.The k-medoids algorithm based on simulated annealing is used to classify line vectors,thus the flexible placement of multi-controllers is realized.Simulation results show that,compared with the KCBP and the optimized K-means algorithm,the proposed algorithm can make the reliability of the control path up to 17% on the basis of guaranteeing the load balancing and low network delay.
-
Computer Engineering. 2018, 44(8): 142-150. https://doi.org/10.19678/j.issn.1000-3428.0049930
Existing methods for trajectory privacy-preserve can’t protect K-anonymous protection from attacks with background knowledge.To this end,a method generating dummy trajectories with user’s true trajectories is proposed.By generating (K-1) dummy trajectories to achieve K-anonymous with true trajectories,adversaries can’t identify them through randomness.By modeling the Markov model of user’s trajectories with considering background information,dummy trajectories can’t be broken because of background information compared with methods of trajectory replacement,trajectory rotation and random walk,the experimental results show its efficiency of dummy trajectories generation and effectiveness of privacy-preserving.
-
Computer Engineering. 2018, 44(8): 151-154,160. https://doi.org/10.19678/j.issn.1000-3428.0047898
Existent heterogeneous cryptosystems don’t have the ability of encrypting any lengths of messages,as well as no ability to sign a signature without knowing the content of the message.Therefore,this paper proposes a hybrid blind signcryption scheme from Identity-based Cryptosystem(IBC) to Certificateless Cryptosystem(CLC).Private Key Generation (PKG) in IBC and Key Generation Center(KGC) in CLC can generate their own system master keys and have the ability to encrypt any long message and blind signcryption.Analysis results show that compared with existent signcryption schemes,the scheme has high speed and efficiency,and small computation,which is more feasible under the random oracle model.
-
Computer Engineering. 2018, 44(8): 155-160. https://doi.org/10.19678/j.issn.1000-3428.0047775
In view of the lack of formal representation of the theft type APT attack,a hierarchical APT representation model is built which named APT-HRM.Referring to the HARM model,the APT attacks are divided into two layers:Attack Chain(AC) and Attack Tree(AT).The AC is composed of 4 stages:reconnaissance,infiltration,exploitation and exfiltration.The AT is composed of attack means in each stage of the AC,and the APT attacks are carried out in stages according to the AC.Analysis results of DUQU 2.0 APT attack show that,the model can effectively describe the APT attack behavior.
-
Computer Engineering. 2018, 44(8): 161-166,173. https://doi.org/10.19678/j.issn.1000-3428.0048395
A high-performance Linux virus detection method is designed using the theory of classification and integration learning in machine learning.This method is,first to train and obtain a number of base classifiers,then to integrate the base classifiers’ classification results to get the final detection result.The method extracts the ELF file feature as the sample feature,and selects BP neural networks as the base classifiers,and uses the improved AdaBoost algorithm to train and integrate the base classifiers.In the Adaboost algorithm,the weight calculation method and integration method of the base classifiers are improved,which makes this algorithm more suitable for the virus detection problem.Experimental results show that the effectiveness of the proposed detection method is verified by relative experiments,which is better than Avria Linux and F-PROT methods.
-
Computer Engineering. 2018, 44(8): 167-173. https://doi.org/10.19678/j.issn.1000-3428.0047650
Cross-site Scripting(XSS) attacks pose serious threats to web applications.Before the application is released,detecting them can effectively reduce the risk of vulnerabilities.Aiming at the problems in the current detection of cross-site scripting,such as missed reports and false alarms,a dynamic detection method is proposed.Based on the basic candidate element library of attack vectors and the initial attack vector seed library,an effective attack vector conforming to the output point type is automatically generated during the detection process.According to the detection result at the current moment,the priority of the attack vector is adaptively adjusted,and all the injection point attacks are performed,after finishing,it traverses the entire site twice to check the vulnerabilities to be discovered.Experimental results show that compared with APPScan、WVS,this method can find more vulnerabilities.
-
Computer Engineering. 2018, 44(8): 174-178,183. https://doi.org/10.19678/j.issn.1000-3428.0047923
In order to ensure that the data stored on the semi trusted cloud server is not tampered or lost,a data integrity verification scheme based on lattices cipher algorithm is proposed.The security of the scheme is based on the LWE difficult problem.The user encrypts the data block by using the homomorphic encryption algorithm and uploads the encrypted ciphertext and the original data to the server.Analysis results show that the scheme can achieve data integrity verification by data holders and support dynamic updating of data.
-
Computer Engineering. 2018, 44(8): 179-183. https://doi.org/10.19678/j.issn.1000-3428.0048127
The Software Defined Network(SDN) provides dynamic control of the network,but the security is poor.Therefore,a dynamic random network model of SDN is proposed to study the process of network virus spreading.Through theoretical analysis and numerical simulation,it is found that the propagation of network viruses from the source subnet to the target subnet is related to the mobility of the nodes between the subnetworks.Experimental results show that the method can effectively identify virus migration threshold and detect the trend of virus transmission in real time.
-
Computer Engineering. 2018, 44(8): 184-191,198. https://doi.org/10.19678/j.issn.1000-3428.0047937
The existing network attack model is not accurate because it does not consider the time cost.Therefore,an attack path generation algorithm based on time cost is proposed to eliminate redundant paths by analyzing the cost weight of attack path.Through the identification of redundant path nodes,the inaccuracy of node confidence calculation is avoided.Experimental results show that the algorithm can effectively eliminate redundant paths,improve the accuracy of node confidence calculation and achieve the optimization of attack path.
-
Computer Engineering. 2018, 44(8): 192-198. https://doi.org/10.19678/j.issn.1000-3428.0047796
In view of the current status of frequent database privacy data leakage incidents,the RSA-based proxy re-encryption authorization access method is designed,and on the basis of it,an optimization method based on a symmetric encryption system is proposed.Construct a permission control layer by encrypting the encrypted data to create a mirroring layer and dividing the user's authority.It can enable data access sharing without copy to prevent data leaks.Experimental results show that compared with the method based on public key system,this method has higher re-encryption and data access efficiency,and improves the access performance of PostgreSQL database.
-
Computer Engineering. 2018, 44(8): 199-203. https://doi.org/10.19678/j.issn.1000-3428.0047712
Most existing directed sensor network scheduling algorithms only focus on isomorphic sensor nodes,without considering the influence of node heterogeneity on the performance of the algorithm.To solve this problem,a heterogeneous nodes scheduling algorithm based on learning automata is proposed.The node scheduling problem is transformed into a set coverage problem,and the probability of the selected perception direction is adaptively updated by using the characteristics of learning automata to construct a number of node coverage sets with satisfying conditions. Simulation results show that compared with greedy algorithm,the proposed algorithm can effectively reduce energy consumption and prolong network lifetime.
-
Computer Engineering. 2018, 44(8): 204-211. https://doi.org/10.19678/j.issn.1000-3428.0047458
For multiple attribute decision making problems with Interval Type-2 Fuzzy Sets(IT2FSs) in the information environment,in this paper a decision model based on risk preference and cross entropy is proposed.Firstly,the relative define of IT2FSs is given,this paper proposes a new ranking value formula.At the same time,the cross-entropy is introduced to distinguish the uncertainty between two IT2FSs,in order to reflect the decision maker’s attitude towards risk,risk preference factor is joined to do as the risk attitude of decision maker.In addition,two models are constructed,which including the attribute weights are completely unknown and the attribute weights are known.By solving the optimal linear programming model to observe the effect of risk preference on attribute weights.Finally,through the analysis of an example to prove the risk preference of decision maker has an effect on attribute weights in this model.
-
Computer Engineering. 2018, 44(8): 212-217,223. https://doi.org/10.19678/j.issn.1000-3428.0047762
Implicit feedback has data noise and lacks negative feedback.When using matrix decomposition method to train recommendation model,the feature extraction is not clear and the recommendation result is skewed.Aiming at these problems,an improved Alternating Least Square(ALS) matrix decomposition online recommendation algorithm with missing data modeling is proposed.The neighbor information is used to select positive samples for the user,at the same time,the negative samples in missing data are modeled according to the popularity of articles.Then the improved ALS algorithm based on elements is combined with online learning.Experimental results on MovieLens data sets show that,compared with eALS and Rcd algorithms,the proposed algorithm can effectively reduce the impact of data noise and matrix sparsity on the matrix decomposition recommendation algorithm,thus improving the accuracy and efficiency of the recommendation.
-
Computer Engineering. 2018, 44(8): 218-223. https://doi.org/10.19678/j.issn.1000-3428.0047107
For defect detection and defect classification problems of Scanning Electron Microscope (SEM) images during wafer inspection,this paper applies a Convolutional Neural Network(CNN) called ZFNet to classify wafer defects.On this basis,a patch-based CNN defect detection algorithm is proposed.For better accuracy and higher speed,another detection algorithm is proposed by modifying Faster RCNN classifier.Experimental results show that,by learning from the defects data marked with locations and types,the two detection algorithms can both detect and classify defects of multiple types on SEM images.
-
Computer Engineering. 2018, 44(8): 224-229,236. https://doi.org/10.19678/j.issn.1000-3428.0047537
For quickly and efficiently find the required semantic Web service from the service register,a semantic Web service discovery method is proposed.The ontology concept is used to describe the semantic Web service and it is transformed into faceted tree.The semantic similarity service is clustered before the service matching,and the service cluster is formed in the service register.Then the number of service comparisons is reduced and the service matching is more efficient.Experimental results show that the precision and recall of the method is up to 87.6% and 90.6%,compared with clustering and two partite graph methods,the service response time is reduced by half,and the efficiency of service discovery is improved.
-
Computer Engineering. 2018, 44(8): 230-236. https://doi.org/10.19678/j.issn.1000-3428.0050502
In order to obtain better Uyghur Named Entity Recognition(NER) performance without the need of resources and relying on artificial features is an important problem to be solved.In this paper,a neural network model based on BiLSTM-CNN-CRF is constructed.Firstly,Convolutional Neural Network(CNN) is used to train character vectors with morphological characteristics such as suffix and prefix of Uyghur words.Then,skip-gram model is used to train large-scale corpus to generate word vectors with semantic information.Finally,a BiLSTM-CRF deep neural network suitable for Uyghur NER is constructed by using concatenated vectors which includes the character vectors,part-of-speech vectors and word vectors as input.Experimental results show that the proposed model can solve the problem of automatic recognition of named entities and has good robustness.Its F1 value reaches 91.89 %.
-
Computer Engineering. 2018, 44(8): 237-242. https://doi.org/10.19678/j.issn.1000-3428.0047077
Aiming at the problem that single feature is hard to adapt to complex scene changes,a multi-feature fusion algorithm is proposed.Based on the Mean Shift(MS) tracking framework,this paper regards the color feature and motion feature characterized by optical flow as two players.By looking for the equilibrium point of the two in the game and as output of tracking the result,this algorithm reaches dynamicallg the optimal balance of different features in tracking process.Experimental results show the the algorithm can adapt to changes of illumination and target’s pose,which has obvious advantages in the case of shelter and interference,compared with traditional tracking methods,this algorithm has higher accuracy and robustness of tracking which can realize excellent tracking in complex scenario.
-
Computer Engineering. 2018, 44(8): 243-250. https://doi.org/10.19678/j.issn.1000-3428.0047261
A method which can animate and simulate the movement of water drop on the surface of an object is put forward to enhance the realism of real-time 3D raining scene.The water drop can only move within the texture space in the existing dimensionality reduction simulation method,so it is divided into two types,which are 2D and 3D water drops,thus realizing the interconversion between the 3D water drop of the world space and the 2D water drop of the texture space.The water drop is regarded as a rigid particle when conducting physical simulation and its equation of movement locus is solved by velocity-verlet integration method.When it is rendered,a height map will be generated based on the simulation data in the texture space,then the liquid bridge phenomenon between the water drops will be simulated based on the Gaussian blur and color gradation algorithm,finally it will be converted to normal map for the rendering of water drops.Experimental results show that this method can simulate the movement of water drop on the surface of an object in real time,which makes dimensionality reduction simulation of the movement of water drop no longer depend on a continuous curved surface.The simulation efficiency of this method is improved by over two and five times respectively when comparing with the Smoothed Particle Hydrodynamics (SPH)-based 2D and 3D simulation method.
-
Computer Engineering. 2018, 44(8): 251-256. https://doi.org/10.19678/j.issn.1000-3428.0048024
The previous works enhance fingerprint images in Fourier domain based on block.But there are some shortcomings in these works,they are not robust enough to non-smooth areas,and the size of the window is not easy to determine.A curved- regions-based fingerprint enhancement algorithm in Fourier domain is proposed to solve these problems.Curved-region transform,which finds curved region and map to 2-D array,is used in Fourier domain,and filters are designed based on the frequency images of curved regions.Fingerprint image is partitioned into smooth areas or non-smooth areas,and only construct curved regions in non-smooth areas in order to improve the execution efficiency.Composite window is introduced to resolve the choice of window size.Experimental result shows that the improved method has better enhancement effect than the algorithms based on block.The matching accuracy of enhanced fingerprints can reach 99.74%.At the same time,the running time is significantly lower than that of the surface Gabor algorithm.
-
Computer Engineering. 2018, 44(8): 257-262,267. https://doi.org/10.19678/j.issn.1000-3428.0047659
When applying most existing object proposal methods on complex indoor scenes,the results show that there are some problems such as ignoring the small size object and objects in planar regions and detection inaccuracies of big objects caused by occlusion.Aiming at above these problems,this paper proposes a layout estimation and object region extraction algorithm for indoor RGB-D scenes.Firstly,it uses the 3D point cloud for plane segmentation and classification.Secondly,it adopts two segmentation methods using RGB-D data for obtaining crude object segments and then utilizes four similarity measures for hierarchical grouping.Finally,based on the results of layout estimation,it takes diversification strategies to fit bounding boxes for different regions.Experimental result shows that the proposed algorithm can improve extraction efficiency obviously and improve bounding box proposal recall score with fewer object candidates.In addition,it does not need pre-training and has fast calculation speed.
-
Computer Engineering. 2018, 44(8): 263-267. https://doi.org/10.19678/j.issn.1000-3428.0048671
The Adaptive Support Weight(SAW) algorithm performs better on accuracy and sensitivity,but it takes too much time on running time.Therefore,an optimized adaptive support weighting method is proposed.It simplifies the ASW algorithm to support the color similarity function in the weight formula,introduces the lleft-right continuity consistency check,finds the occluded point in the initial disparity map and performs occlusion filling.The median filter removes the horizontal stripes that may appear in the occluded filling and smooths the image.Experimental results show that the method can reduce the average running time of ASW algorithm by 20% and the average false match rate by 2.76% when the support window is unchanged,and has good robustness to illumination.
-
Computer Engineering. 2018, 44(8): 268-272,278. https://doi.org/10.19678/j.issn.1000-3428.0047121
Aiming at the problem of static image human behavior recognition,a method of merging Local-constrained Linear Coding(LLC) and global feature descriptors is proposed.This method makes intensive sampling of the image,and extracts the Scale-Invariant Feature Transform(SIFT) features in each subregion.Then the LLC method is used to encode and pool the dense SIFT features.In order to add spatial information,spatial pyramid information is used to extract LLC pooling features with spatial location information.Generalized Search Trees(GIST) features are extracted,and the final features of the image are described using LLC pooling features tandem GIST features.The kernel function is the histogram cross kernel function Support Vector Machine (SVM) and it is used for classification.Experimental results show that compared with the methods using LLC,Spatial Pyramid Matching(SPM) features and GIST features,the proposed method has better recognition effect.
-
Computer Engineering. 2018, 44(8): 273-278. https://doi.org/10.19678/j.issn.1000-3428.0048088
In order to adapt to the high-speed network environment and realize the accurate measurement of network traffic,a network algorithm of flow equal probability sampling based on the combination of Counting Bloom Filter and packet-based flow sampling technique is proposed.It identifies whether there is a new flow by the 4 bit Counter vector,and realizes the equal probability sampling of the network flow by adjusting the sampling frequency constantly to compensate for the error rate of the new flow judgement.Then,it makes an equal probability sampling of network flow and obtains a more realistic distribution of network traffic.Experimental results show that the algorithm’s measurement results are close to the real value of network flow,and it has scalability,which can meet the current demand of traffic measurement in the complex and changeable high-speed network environment.
-
Computer Engineering. 2018, 44(8): 279-284. https://doi.org/10.19678/j.issn.1000-3428.0047817
Most existing haptic interaction studies focus on simulating the haptic interaction of single-phase flow,but because they ignore the differences between the different liquids in miscible liquids,they cannot be directly applied to the haptic interaction of miscible liquids.In order to solve the problems such as unreality and instability in the process of miscible liquid haptic interaction,a new haptic rendering method is proposed.Using this method,real and stable haptic forces can be obtained through calculation during liquid mixing.In addition,the calculation of the tactile force of the friction force during the calculation of the haptic force can improve the reality of the haptic interaction.Experimental results show that compared with the traditional miscible liquid haptic interaction method,this method can improve the reality and stability of the haptic force,thus providing the user with a more realistic experience.
-
Computer Engineering. 2018, 44(8): 285-290. https://doi.org/10.19678/j.issn.1000-3428.0047170
In Wireless Sensor Network(WSN) with Radio Frequency(RF) energy scavanging,Energy Providing Node(EPN) and Energy Harvesting Node(EHN) are needed to deploy reasonably to transmit data and prolong network lifetime.A data transmission scheme based on WSN technology with RF energy scavanging in smart grid is presented,as well as a data aggregation tree routing algorithm with prior-recharging mechanism.At least one EPN is arranged around the EHN.Energy-efficient data transmission can be achieved through Integer Linear Programming(ILP) and least-weight-scavanging spanning tree algorithm.Some heavy load EHNs are choosed for prior recharging in network operation in the finally built tree.Simulation results show that the scheme has smaller algorithm complexity,and can extend the network lifetime effectively while completing data transmission.
-
Computer Engineering. 2018, 44(8): 291-295. https://doi.org/10.19678/j.issn.1000-3428.0047410
Aiming at the lower matching precision problem of existing map matching algorithms in intensive road,this paper puts forward an improved matching algorithm with curve fitting,which puts relative error as the benchmark.Firstly,the interpolation method is used to complete the anomaly positioning points.Then the road topological structure is used to determine the collection of retrieval roads.Finally,the principle of relative error and minimum is used to fit the road straight line and select the vehicle traveling road,and the vertical projection method is used to determine the road position of the vehicle.Compared with the Beidou data samples,the matching rate of the proposed algorithm is more than 90%,and the single point average matching time is less than 5 ms.Compared with the traditional curve matching algorithm,the accuracy and real-time performance of this algorithm can be improved,and its application is stronger,especially in large cities with complex roads.
-
Computer Engineering. 2018, 44(8): 296-300. https://doi.org/10.19678/j.issn.1000-3428.0048106
The wide bandwidth of wide-band Linear Frequency Modulation (LFM) signals will cause pressure on traditional data processing.To slove this problem,a method of 2-D DOA estimation for wide-band LFM signals with spatial under-sampling is proposed.Based on the energy aggregation characteristic of the Fractional Fourier Transform(FRFT) on the wide-band LFM signals,the signals source is separated into a stationary single frequency signal,and a new spatial time frequency data model is established.On this basis,a spatial Uniform Circular Array(UCA) is constructed to process the incident signal,and then the integer search method is used to estimate the pitch angle and azimuth of the received signal. Simulation results show that compared with the array form of non-uniform L arrays,the proposed method can effectively improve the estimation accuracy,reduce computation complexity and computational complexity.
-
Computer Engineering. 2018, 44(8): 301-307,314. https://doi.org/10.19678/j.issn.1000-3428.0047642
A speed planning method based on intermediate speed is proposed for intelligent vehicles to meet the need of speed planning and safe parking when facing obstacle during the process of autonomous driving.It introduces the speed of the vehicle obstacle into the planning process and takes the intermediate speed as reference quantity of speed control .By this way,the advance planning of vehicle speed is realized.Furthermore,it solves the problem of fluctuation of speed planning process when facing the dynamic obstacle.The proposed method is simulated on the CarSim/Simulink co-simulation platform,and the experimental results show that it can improve the speed stability of intelligent vehicle,and has good adaptability to the obstacle.
-
Computer Engineering. 2018, 44(8): 308-314. https://doi.org/10.19678/j.issn.1000-3428.0047499
Radio Frequency Identification(RFID) technology uses radio signals to identify tags automatically.The recognition does not need objects to be in line of sight,which can overcome the limitation of the traditional bar-code or visual-based approaches.So this paper focuses on a fast approach for dynamic object localization based on the fusion of the RFID phase and the laser information.To be precise,it achieves the velocity estimation of an object attached with an RFID tag by the phase difference.On this basis,it uses a laser range finder to estimate speed,angle and distance of the obstacles.Then it selects the best match point between RFID velocity and laser velocity,which is used to determine the location of the object.Finally,it achieves the localization of dynamic objects based on the angle and distance information of laser.Experimental results show that the proposed method can effectively improve the accuracy of localization.
-
Computer Engineering. 2018, 44(8): 315-320. https://doi.org/10.19678/j.issn.1000-3428.0050269
Modern commercial multicore processors lack a hardware-supported mechanism for processing inter-core communication,and multiple processing cores must pass data through locked shared memory.Therefore,a software-based lock-free queue is designed as an inter-core communication mechanism,to improve the performance of the software queue through the data structure without lock.When the data arrival rate is lower,the queue adaptively reduces the queue length,thus taking up less memory space and making better use of processor cache;When the data arrival rate is higher,the queue adaptively increases the queue length to avoid data loss.Experimental results show that this queue has higher data processing performance than FastForward,MCRingBuffer and B-Queue queues in practical application scenarios where the data arrival rate changes greatly.
Volume 50 No 4
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