15 November 2017, Volume 43 Issue 11
-
Select all|
-
Computer Engineering. 2017, 43(11): 1-7. https://doi.org/10.3969/j.issn.1000-3428.2017.11.001Considering the benefits of edge computing and the demand of smart home,this paper applies edge computing to smart home environment at the premise of saving computing resources and improving computing resources utilization.A system based on personal computer is constructed to provide edge computing service for smart home environment.By using existing technologies such as virtualization and container,this system can converge cloud services on the personal computer platform,acting as a local computing unit to provide edge computing service for users.And energy-saving and scheduling algorithm is designed to implement system energy-saving management by making service management programs and smart gateway work together.Test results show that the system is easy to deploy and manage and it can provide better service quality and scalability with limited computing resources compared with simply using cloud computing system only.
-
Computer Engineering. 2017, 43(11): 8-15. https://doi.org/10.3969/j.issn.1000-3428.2017.11.002In order to realize the efficient processing of large scale and dynamic partition information in the big data environment,a real-time database partitioning system is proposed combining with the flow computing framework.This system copes with the large scale and dynamic workloads by using stream computing technologies in the big data environment.It designs a real-time data partitioning algorithm to realize automatic and immediate generation of data partitions.The system realizes the scalability and high-throughput adaption by using the horizontal scaling mechanism of streaming computing framework.The experimental results show that the system can realize efficient and real-time database partition in big data environment.It has higher partitioning quality and lower time than tranditional partitioning algorithm.
-
Computer Engineering. 2017, 43(11): 16-21,26. https://doi.org/10.3969/j.issn.1000-3428.2017.11.003The rapid increase in the amount of data have driven many database systems to provide nearlinearly scalability,but most of these systems cannot automatically maintain the strong consistent relation among objects.Aiming at this problem,this paper designs a distributed transaction processing mechanism with high scalability.For the semantic and storage characteristics of information network model,it dynamically controls the group commit size by the association between received transactions to use object caching and improve system throughput.Meanwhile,it concurrently executs the same operations between storage node and its replications rather than replicating with log,so as to improve the response speed.Experimental result shows that,compared with two-phase commit protocol,the proposed mechanism can obtain higher scalability which approximates to linear while ensuring strong transactional characteristic.It also has great improvement in system throughput and response time.
-
Computer Engineering. 2017, 43(11): 22-26. https://doi.org/10.3969/j.issn.1000-3428.2017.11.004Existing wait-free universal construction algorithm only considers the bounded wait-free situation and can not be adapted to unbounded wait-free model.This paper presents a novel solution:Unbounded Wait-free Universal Construction(UWUC for short) algorithm which uses Column Address Strobe(CAS) object and Fetch&Add object.The number of processes arrived during a time interval is finite,thus using a special queuing technical and helping mechanism implementing the unbounded wait-free universal construction.Experimental results show wait-free characteristics of UWUC algorithm.
-
Computer Engineering. 2017, 43(11): 27-31,39. https://doi.org/10.3969/j.issn.1000-3428.2017.11.005The effccts of music melody and beat on Electroencephalogram(EEG) wave energy spectrum hasve great application significance.This paper proposes a double simulation inducing method with light music and music beats in the hearing.It regulates the brain activity of subjects,and by analyzing the collected EEG signal to determine whether can quickly and effectively enter the highly relaxed state.In the regulating stage,subjects are asked to listen the mix of light music and music beat with the gradual speed and the opposite power order for 5 min,and analyses the signals collected by the EEG headset.Experimental results show that when the duration of regulation is about 2 min~3 min,the method based on double stimulation of light music and music beats can effectively promote subjects to enter the highly relaxed state.
-
Computer Engineering. 2017, 43(11): 32-39. https://doi.org/10.3969/j.issn.1000-3428.2017.11.006
Aiming at the problem of low utilization ratio of the channel in the Time Division Multiple Access(TDMA) mechanism in the tree topology underwater acoustic sensor network,this paper proposes a hybrid MAC protocol based on improved lightweight traffic adaptive random medium access mechanism and TDMA mechanism:TAST-MAC.It adopts S-ALOHA competition mechanism effectively to improve the channel utilization when the network traffic is low.This protocol uses an optimization algorithm to calculate the maximum transmission probability table values which are assigned to the cluster sub nodes,and uses a flexible traffic adaptive slot allocation mechanism to adapt different network traffic situation.Experimental results show that TAST-MAC not only improves the communication efficiency in the tree topology networks,but also reduces the waiting time of the sub nodes.
-
Computer Engineering. 2017, 43(11): 40-44. https://doi.org/10.3969/j.issn.1000-3428.2017.11.007Detection results of the fault detection method based on machine learning or Bayesian network are only based on the inference of probability or historical information.Therefore,it cannot be applied to the fault detection and automatic repair system.Aiming at this problem,this paper proposes a fault detection method based on production rules.It makes the known failure be automatically issued and the unknown failure be notified to the network administrator through alarm form.Meanwhile,it proposes a self-adaptive polling algorithm to further reduce the overhead of the system without losing the response time of the fault.Experimental result shows that the proposed method can reduce polling overhead and control message exceptions effectively.
-
Computer Engineering. 2017, 43(11): 45-49. https://doi.org/10.3969/j.issn.1000-3428.2017.11.008Aiming at the problem that too many antennas of large-scale Multiple Input Multiple Output(MIMO) system increases the amount of calculation of precoding algorithm,this paper adopts the method combining antenna selection algorithm with traditional precoding algorithm in Coordinated Multiple Points(CoMP) transmissionsystem,through reducing the interference between zones and interference between users,it can avoid the higher complexity brought by traditional precoding algorithm to channel matrix inversion with the increase of antenna numbers.Theoretical analysis and simulation results show that compared with Zero Forcing (ZF) precoding algorithm,Minimum Mean Square Error(MMSE) precoding algorithm and precoding algorithm based on maximization of signal to leakage plus noise ratio,antenna selection and joint precoding algorithm can better improve the channel capacity of the system and reduce the complexity of precoding algorithm.
-
Computer Engineering. 2017, 43(11): 50-54. https://doi.org/10.3969/j.issn.1000-3428.2017.11.009To better balance system throughput and user fairness of LTE-A systems,a resource management based on greedy algorithm is proposed.The Channel State Information(CSI) information reported by the client is fully utilized to reduce the performance impact of carrier aggregation by channel variation,and the greedy algorithm is adopted to solve the NP-hard problem of resource allocation.In scheduling process,the weight factor is redesigned by using occupied resource blocks and average rate to guarantee the different Quality of Service(QoS) requirements of all kinds of users.Simulation results show that compared with LL+PF algorithms,the proposed algorithm has better performance both in user fairness and system throughput.
-
Computer Engineering. 2017, 43(11): 55-65. https://doi.org/10.3969/j.issn.1000-3428.2017.11.010
Vehicular Ad Hoc Network(VANET) single-layer construction routing protocols consider less factor,causing low packet delivery ratio and high average end to end delay.Considering the influencing factors of vehicles position,speed,density of intersections,wireless link quality,and MAC layer frame error rate,this paper proposes Multi-factor Cross-layer Position-based Routing Protocol for VANET(MCLPR) for urban environment.An algorithm for Vehicle Selection at Intersections(AVSI) is designed to extract the cross-layer information of the wireless link quality and the MAC layer frame error rate.The weight value of each influencing factor is calculated by using the Analytic Hierarchy Process(AHP) to determine the best forwarding path.Simulation results show that,compared with routing protocols such as AODV and DSDV,MCLPR protocol has higher packet delivery ratio and lower end to end delay,ensuring the reliability and efficiency of data transmission.This routing protocol is suitable for urban environment with the large network load and density,fast moving speed.
-
Computer Engineering. 2017, 43(11): 66-69,75. https://doi.org/10.3969/j.issn.1000-3428.2017.11.011In order to solve the problem of imbalance between the estimation accuracy and estimation range of the intercepted signal with low Signal-to-Noise Ratio(SNR) in electronic environment,such as radio monitoring and electronic warfare and other non-cooperation communications,this paper proposes an improved Non-Data-Aided(NDA) frequency shift estimation algorithm.The influence of the maximum correlation interval to estimation range is removed through the difference function of the self-correlation of Q times of the normalized baseband signal.Then an effective estimation value is accomplished by the weighted summation of the difference phase.Simulation results show that the proposed algorithm can obtain the estimation accuracy of almost 10-4 even if the SNR is low to-15 dB,which could meet the requirements of the actual monitoring.
-
Computer Engineering. 2017, 43(11): 70-75. https://doi.org/10.3969/j.issn.1000-3428.2017.11.012In order to improve the balance of load management in Content Distribution Network(CDN),this paper proposes a fast control packet super gradient dual decomposition algorithm for CDN load management.The paper presents a simplified DNS control load management model in CDN,which is sufficient to solve the basic operational problems faced by CDN operators.The convex optimization model of CDN load management problem is formulated,and a distributed dual algorithm is proposed to solve the convex optimization model and implement performance improvement.Simulation experimental result shows that,compared with the OSPR algorithm and GS algorithm,proposed algorithm has lower data transmission delay and energy consumption
-
Computer Engineering. 2017, 43(11): 76-80. https://doi.org/10.3969/j.issn.1000-3428.2017.11.013In order to extend the network life on the base of lower Symbol Error Rate(SER),a relay selection scheme based on Hybrid Decode-Amplify-and-Forward(HDAF) relaying method is proposed.Joint optimization function about receiving Signal to Noise Ratio(SNR) at the destination and network life is defined to select the optimal relay.For the given threshold SNR at the relay nodes,all relay nodes are divided into two types:forwarding signal by Amplify-and-Forward(AF) relaying method and forwarding signal by Decode-and-Forward(DF) relaying method.The receiving SNR at the destination for each relay node is calculated and the optimal relay is selected reasonably according to the joint optimization function.Simulation results show that under the same conditions,the proposed relay selection scheme,compared with the relay selection scheme based on AF relaying method,not only achieves much better performance in SER,but extends the network life,improves system performance.
-
Computer Engineering. 2017, 43(11): 81-89. https://doi.org/10.3969/j.issn.1000-3428.2017.11.014In order to solve the current Wireless Sensor Network(WSN) source localization algorithms to synchronize and optimize time delay estimation and angle estimation,and it cannot segment noise and signal subspace,this paper proposes an improved source precise localization algorithm for WSN,it uses circulation mode to structure signal analysis mechanism,and implements the real time analysis segmentation of signal space in frequency domain,the noise of the signal subspace and the signal subspace is divided into matrix signal independent,the time delay estimation and the angle estimation of the source precise positioning are gained.Based on energy spectrum density estimation,orthogonal multiplexing cycle mechanism is designed,signal eigenvalue is decomposed,precise estimation of positioning signal features is achieved,the precision of the time delay and angle estimation is improved,and the signal subspace with minimum time delay estimation and the minimum angle estimation of space from the collection is filtered out ,thus the algorithm completes the unified time delay estimation and angle estimation,improves the accuracy of source localization process.Simulation experimental results show that compared with the traditional DT-IPL algorithm,CD-CPP algorithm,this algorithm can significantly improve the positioning accuracy and the deviation between the source position and the actual position is low in high decline channel condition.
-
Computer Engineering. 2017, 43(11): 90-96,101. https://doi.org/10.3969/j.issn.1000-3428.2017.11.015In underwater sensor network,most existing repair algorithms of network topology just consider to repair the connectivity of the network,without taking into account the network lifetime.To deal with the above problems,an algorithm based on redundancy node selection model is proposed.In this algorithm,a distributed manner is adopted to select critical nodes after network deployed,in which the redundancy nodes are selected to complete the topology repair by moving to the position when the critical nodes are failure.Meanwhile,the sleep/wake strategy is adopted to prolong the network lifetime.Experimental result shows that,compared with Block Movement Repair(BMR)algorithm,the proposed algorithm can greatly reduce the node’s moving distance,prolong the network lifetime and guarantee delivery ratio.
-
Computer Engineering. 2017, 43(11): 97-101. https://doi.org/10.3969/j.issn.1000-3428.2017.11.016This paper analyzed two different ways of the security of SPN structure,16 different optimal S-Boxes classified and permutation equivalence class whose representative elements are the 16 elements mentioned above.It evaluates the two different results by two cryptographic properties:Maximum Everage Differential Probablities(MEDP),Maximum Everage Linear Hull Probablities(MELP).Analysis results show that the permutation equivalence class as the confuse layers are more concertrated,which measns has a better ability to resist differential attack than those constructed by using affine equivalence class in MEDP,while the other cryptographic properties make no difference in MELP.
-
Computer Engineering. 2017, 43(11): 102-107. https://doi.org/10.3969/j.issn.1000-3428.2017.11.017Aiming at realizing the demand for local refinement in video encryption,a scheme of selective video encryption based on Largest Coding Unit(LCU) is proposed.Combined with High Efficiency Video Coding(HEVC),it obtains the difference charts of CU on quad-tree partition.It designs a LCU Correlation Distance(LCD) model based on the center distance weight and gray correlation of LCU to count the LCU correlation of all the difference charts,so that all LCUs have selective encryption capability.Arnold scrambling algorithhm for the LCU block can be directly carried out and encrypting multi LCU blocks makes the decryption greatly difficult.Experimental results show that this scheme has a really safe encrypting capability for the video key areas compared with existing videl encryption schemes.
-
Computer Engineering. 2017, 43(11): 108-116. https://doi.org/10.3969/j.issn.1000-3428.2017.11.018Collusion attack produced a serious threat to e-commerce trust evaluation mechanism and the existing attack detection model is not able to solve the problem well,combined with the characteristics of domestic C2C e-commerce,this paper uses the goods in collusion attack as the object of detection,fuses explicit and implicit user behavior feature and summary feature of related transactions and buyers,combining the advantages of DS evidence theory in dealing with uncertain problems,puts forward the collusion attack detection model under DS evidence theory.Experimental results based on the real collusion attacks data in e-commerce show that the proposed reasoning model can effectively identify the collusion attack,the extracted attacks can effectively reflect the user’s real behavior,effectively distinguish between the goods involved in attacks and legitimate transactions.
-
Computer Engineering. 2017, 43(11): 117-121. https://doi.org/10.3969/j.issn.1000-3428.2017.11.019
Khudra algorithm is a kind of lightweight block cipher algorithm which has 18 rounds.The existing analysis method,which uses the impossibility relevant key difference to analysis Khudra algorithm,constructs a 14 rounds distinguisher to attack the 16 round Khudra algorithm by introducing a difference on two keys.The successful attack probability of the distinguisher is 2－56.85.In this paper,it constructs 10 rounds distinguisher,by making difference on one key,to attack Khudrain total 16 rounds based on the relevant key difference.Analysis results show that the successful probability of the 10 rounds distinguisher improvs 2 28.425 compared with the previous 14 rounds distinguisher,data complexity of the whole analysis process is 2 33,and time complexity is 2 95.
-
Computer Engineering. 2017, 43(11): 122-127,139. https://doi.org/10.3969/j.issn.1000-3428.2017.11.020Aiming at the phenomenon that there is an increasing number of repackaged applications in the Android third-party application market,this paper proposes a method of detecting Android repackaged applications by using function call graph.It decompiles the application to gain the Smali code,analyzes the Smali code to generate a function call graph,processes the function call graph by using theoperationcode as the attribute of the node,filters the third-party library,and saves the Application Program Interface(API) associated with the interface.On this basis,it uses the Motifs’ substructure to representthe function call graph.According to the similarity of the subgraph,it computes the similarity of the application,so as to determine whether it is a repackaged application.Detection results in 1 630 applications in the market show that the proposed method has higher accuracy and better expansibility.
-
Computer Engineering. 2017, 43(11): 128-133,145. https://doi.org/10.3969/j.issn.1000-3428.2017.11.021In view of the shortcoming of failure in cross-domain anonymous authentication scheme based on certificateless public key cryptography,a cross-domain authentication scheme in cloud environment is proposed by using bilinear mapping.On the basis of the validity of certificateless signature and the legitimacy of the message,the proposed scheme can identify the authenticity of user and cloud service provider,and the negotiation of the session key is completed in the bidirectional cross-domain authentication process.Furthermore,the introduction of hierarchical ID tree structure ensures the uniqueness of identity,and the two-factor authentication process of "password + key" strengthens the security of the cross-domain authentication scheme.The scheme realizes the anonymity of user identity by using temporary identity and has controllability for anonymous malicious behavior.Analysis results show that the proposed scheme is secure in CK model,it can resist forgery attacks,replay attacks and replace attacks,and has higher safety performance.
-
Computer Engineering. 2017, 43(11): 134-139. https://doi.org/10.3969/j.issn.1000-3428.2017.11.022Non-negative matrix does not consider the manifold of data when represents multi-view data,which results in the ineffective express of the data internal expression.In this paper,Hessian regularized Non-negative Matrix Factorization(NMF) is proposed.By using the L2 model of Hessian functional,the local topology of the sample is preserved and the algorithm is further extended into Hessian Regularized Joint Non-negative Matrix Factorization(HR-J-NMF) to work on multi-view data.Experimental results show that the Hessian regularized NMF and the HR-J-NMF have a great improvement in both clustering accuracy and mutual information value.The performance of the two algorithms is superior to that of the traditional NMF algorithm.
-
Computer Engineering. 2017, 43(11): 140-145. https://doi.org/10.3969/j.issn.1000-3428.2017.11.023Proposed node cooperation incentive mechanisms are always based on historical transaction information of nodes.Besides the considerable costs of information storage and process,there also exists fraudulent transaction information provided by malicious nodes.Aiming at these questions,it establishes a cooperation incentive mechanism based on adjustments of node property.Nodes adjust the number of their performed transactions depending on being a winner or loser in the learning process of strategies.Simulation results show that the fraction of cooperators significantly increases compared to the original scenario in which the mechanism is not used,and the cooperator fraction has good stability in the equilibrium state when disturbances exist in node strategy selection process.
-
Computer Engineering. 2017, 43(11): 146-151,160. https://doi.org/10.3969/j.issn.1000-3428.2017.11.024The traditional Collaborative Filtering(CF) algorithm mines user’s interest unilaterally according to scoring of resources,none of dividing the user’s interests and considering the user’s interest in time,and these affect the quality of recommendation.In order to solve the above problems,this paper proposes a recommendation algorithm which adapts to changes in user interests and exploits user interests effectively.The algorithm considers the effect of tags based on the traditional CF algorithm and divides the user interest by tags clustering and does the dual choices of similar users in the two aspects of the tags and scoring of resources.Considering change of user interest in the passage of time,the paper puts the time factor into the calculation of the weight of the tags and the scores in order to modify on the time dimension.Experimental results show that the improved algorithm can mine user better and adapt to changes in the user’s interest and improve the accuracy of the recommendation.
-
Computer Engineering. 2017, 43(11): 152-160. https://doi.org/10.3969/j.issn.1000-3428.2017.11.025Diversification retrieval is used to solve users’ information needs,which typically described by query phrase are often ambiguous and have more than one interpretation.This paper researches microblog diversification retrieval,and proposes a novel microblog diversification retrieval method,diversification learning to rank method is applied to microblog diversification retrieval.It develops a series of social media features considering the characteristics of microblog and subtopics distribution,and adds these features one by one to the baseline model which only considering the relational features and the text diversity feature to verify the effectiveness of them.Experimental results show that diversification learning to rank approach can solve microblog diversification retrieval problem,and improve the effectiveness of microblog retrieval.
-
Computer Engineering. 2017, 43(11): 161-165,172. https://doi.org/10.3969/j.issn.1000-3428.2017.11.026Spectral clustering eigenvector of the Laplace matrix is not limited to the distribution shape of the original data and can converge to the global optimal solution,but it cannot accurately reflect the actual relationship between samples.However,fuzzy kernel clustering can use fuzzy mathematics theory to determine the fuzzy relations among samples.For this purpose,this paper merges the fuzzy kernel clustering into spectral clustering algorithm and puts forward SC-KFCM algorithm on the basis of the adjustment of a similarity measure function and distance measurement function,which can make up for the impact that hard part in spectral clustering brings to the clustering results.It uses fuzzy partition to improve hard part in spectral clustering through establishing fuzzy subordinate relations and utilizes the degree of similarity and correlation between eigenvector among clustering samples to improve the hard part in spectral clustering.The experimental results prove that SC-KFCM has more stable clustering results and higher clustering accuracy on data utilizes sets of different distribution characteristic and different dimensions.
-
Computer Engineering. 2017, 43(11): 166-172. https://doi.org/10.3969/j.issn.1000-3428.2017.11.027For attribute reduction problem in inconsistent decision table whose attribute values are three-parameter interval grey numbers,two attribute reduction algorithms are presented.According to the preference between grey information,a grey dominance relation is constructed based on the dominance extent of two grey numbers.A reduction algorithm called up-down approximate reduction based on discernibility matrix is proposed with considering the comprehensive influence of incompleteness and inconsistence.In order to reduce the complexity of reduction process,two kinds of attribute importance are defined.A heuristic reduction algorithm based on two kinds of attribute importance is put forward.Experimental results illustrate the proposed algorithm could not only cope the attribute reduction in inconsistent decision table whose attribute values are three-parameter interval grey numbers,but also possess a lower algorithm complexity when compared with the algorithm based on one kind attribute importance.
-
Computer Engineering. 2017, 43(11): 173-181. https://doi.org/10.3969/j.issn.1000-3428.2017.11.028According to the highly uncertainty group judgment complexity environment bring by Failure Mode and Effects Analysis(FMEA),this paper proposes an novel Interval Type-2 Trapezoidal Fuzzy(IT2 TF) group judgment aggregation method to compute the centralized tendency and fluctuation range of the membership function through the deep mining data information.Based on it,the IT2 TF membership function parameter model is obtained.The model aggregates a set of interval individual judgment with uncertainty information into IT2 TF group judgment result,thereby quantifies the group judgment consensus as well as the uncertainty information by integration of individual judgment.The practicability and effectiveness of the proposed method is demonstrated by an illustrative example and comparative analysis,and indicates it can improve the accuracy of the gourp judgement result.
-
Computer Engineering. 2017, 43(11): 182-186. https://doi.org/10.3969/j.issn.1000-3428.2017.11.029Traditional LFM community discovery algorithm based on local information of network to realize community detection.However,the structure information in the whole network is underused.It leads to serious decline of algorithm precision in network with fuzzing community structure.What’s more,because of community expansion by local information,it’s easy for LFM to get abnormal community division.In order to solve the above problem,an improved LFM algorithm is proposed.Random walk theory is used to measure the similarity of nodes,so that the community structure is clearer.Meanwhile,in order to avoid abnormal community division,maximal clique in the weighted network is found and used to expand the community.The experimental results on artificial networks and real networks show that the improved LFM algorithm achieves higher classification accuracy than traditional LFM algorithm and Label Propagation Algorithm(LPA).
-
Computer Engineering. 2017, 43(11): 187-192. https://doi.org/10.3969/j.issn.1000-3428.2017.11.030
Aiming at the problem that the recognition rate of traditional handwritten digits recognition method is low,this paper proposes a Fused Convolutional Neural Network(F-CNN) model.By combining the high-level features of the Siamese Network(SN) model and Binary Convolutional Neural Network(B-CNN) model,the F-CNN model expands the size of the high-level layers and enhances the features-expression ability of deep CNN network model.In the process of network training,a kind of periodic data shuffle strategy is designed to improve the convergence rate of the F-CNN model to realize better handwritten digits recognition.Experiments results on the public MNIST dataset show that the proposed F-CNN model has 99.10% recognition rate for handwritten digits,which outperforms the SN model and the B-CNN model.
-
Computer Engineering. 2017, 43(11): 193-197. https://doi.org/10.3969/j.issn.1000-3428.2017.11.031Recommendation algorithm based on tensor factorization has low accuracy and data sparseness problem.Therefore,on the basic of the traditional tensor decomposition model,this paper introduces the user nearest neighbor information,and proposes N dimensional tensor decomposition model.Using context aware information,it uses implicit feedback information as the third dimension to establish N dimensional tensor decomposition model.To further improve the the quality of recommendation,it adds the user nearest neighbor information to optimize the N dimensional tensor decomposition model to improve the accuracy of the tensor decomposition recommendation algorithm.Experimental results show that the tensor decomposition recommendation algorithm combined with user nearest neighbor has better accuracy than the traditional tensor decomposition algorithm,can effectively solve the sparsity and accuracy problems.
-
Computer Engineering. 2017, 43(11): 198-203. https://doi.org/10.3969/j.issn.1000-3428.2017.11.032In the big data environment,the computer core technology of public security organs is insufficient,many alternative suspects and the forecast method is relatively backward problem,aiming at these problems,this paper proposes the model of using the Support Vector Machine(SVM) to predict the suspect.According to the historical crime record,the model carries on the feature selection firstly,based on SVM training the suspect characteristic prediction model.Through this model,it can predict the various characteristics of the suspect,and calculate the similarity between the characteristics of prediction and the staff characteristics of the suspect database,then can predict the most possible suspects.Experimental results show that compare to the previous models which using classification and regression algorithm,this model has a good explanation for the prediction results; Besides,it can narrow the scope of the investigation.
-
Computer Engineering. 2017, 43(11): 204-209,215. https://doi.org/10.3969/j.issn.1000-3428.2017.11.033In order to improve the accuracy of automatic registration of three-dimensional scattered point clouds,a new registration algorithm on the basis of region expansion is proposed.Through variations of the normal vector of local point-cloud,characteristic points are extracted.By applying the regional expansion method for initial registration,mismatches can be directly eliminated while searching for accurate matching points.Then,the advanced closest point iterative algorithm is used to accurately align point-clouds.According to experimental results,compared with the classic closest iterative point algorithm and curvature-based point-cloud automatic registration algorithm,this algorithm is capable of improving the accuracy of point cloud registration and has good effects towards the point-cloud model of feature smoothing.
-
Computer Engineering. 2017, 43(11): 210-215. https://doi.org/10.3969/j.issn.1000-3428.2017.11.034For the random images,there are problems of redundant and blur image information,which cannot meet the technical requirements in terms of feature detection,target recognition and 3D reconstruction,based on image quality constraint,this paper proposes a disordered image key frame extraction method.The clustering algorithm without presetting K-means is used for automatic clustering of disordered images.According to the similar distance,the frame,being closest to the clustering center from each clustering,are extracted as the key one.The quadratic blur processing algorithm is adopted to evaluate the extracted key frames for non-reference image quality.If the evaluation values are in accordance with quality requirements,the frames will be reserved.Otherwise,returns to the original cluster,extracts and evaluates the key frame again until the extracted frames meet the quality requirements.Experimental results show that the proposed method can filter out redundant images and extract the key frames satisfying the quality requirements.
-
Computer Engineering. 2017, 43(11): 216-220,226. https://doi.org/10.3969/j.issn.1000-3428.2017.11.035The fog removal method based on fog image imaging model exists distortion phenomenon in the sky region and the calculation of transmissivity at the edge is not accurate.In order to solve this problem,this paper proposes an image dehazing method.The method constructs energy functional which consists of data term,smoothness term,and edge-preserving term based on variational model.The energy functional is minimized by the gradient descent flow method to achieve accurate transmissivity.According to the acquired transmissivity values and fog image restoration theory,the accurate restoration of the image is realized.Experimental results show that the dehazing effect of the proposed method is superior to that of the traditional method in the sky region and the edge region,and has smaller mean square error value and greater structural similarity value.
-
Computer Engineering. 2017, 43(11): 221-226. https://doi.org/10.3969/j.issn.1000-3428.2017.11.036
As the Aggregate Channel Features(ACF) algorithm has many false detection windows,a coarse-to-fine cascaded pedestrian detection algorithm is proposed.ACF is employed as the coarse detector,and then the channel features are improved to filter out false detection windows.The Principal Component Analysis(PCA) filter bank are learned from each channel,instead of learning filter bank from the training image and convolution maps.The single-layer convolution is executed on the filter bank and channels,to reduce feature dimension and improve feature discrimination capability,Finally,a pooling operation is applied on convolution maps to reduce the feature dimensionobtaining improving channel features.Simulation results show that compared with the original ACF algorithm,the proposed method has less false detection windows and the detection rate on INRIA and Caltech databases increases by 3.8% and 17.5% respectively.
-
Computer Engineering. 2017, 43(11): 227-233,238. https://doi.org/10.3969/j.issn.1000-3428.2017.11.037In facial expression generation,most of traditional methods are difficult to produce fine facial expression by linear combination,and the images need to be obtained with 3D laser scanner,which is not conductive to the promotion of the method.In order to solve this problem,a method Non-uniform Rational B-spline(NURBS) building 3D face expression model is proposed.In this method,the control points of the surface are adjusted according to the feature points related to facial expression,and geometric constraint and smoothness constrain are to introduced to guarantee global and local smoothness.Experimental results show that compared with the linear method,this method is more robust,can easily generate fine 3D facial expression from standard image and unconstrained image.
-
Computer Engineering. 2017, 43(11): 234-238. https://doi.org/10.3969/j.issn.1000-3428.2017.11.038In order to apply Convolutional Neural Network (CNN) to video understanding,a recognition algorithm based on trained image CNN features is proposed.Image RGB data is employed to recognize human action in videos.Off-the-shelf CNN models are used to extract features from images,and classification is made by recurrent neural networks with Long Short-Term Memory (LSTM) unit.The research focuses on the choice of CNN architectures and layers,feature vectorization and dimentionality reduction.Experimental result shows that the algorithm has higher accuracy than attention model algorithm and composite LSTM algorithm using RGB data.
-
Computer Engineering. 2017, 43(11): 239-244,251. https://doi.org/10.3969/j.issn.1000-3428.2017.11.039In order to solve the carrier source mismatch problem in image steganalysis,a universal unsupervised steganalysis forensics algorithm combining image retrieval and outlier detection is proposed.For the measured image,The carrier images with the same statistical properties are retrieved from the image database as aided samples.The dense image can be regarded as the outlier in the carrier images.The unsupervised general blind detection of dense images is realized by means of anomaly detection,which avoids mismatch problems and complex classifier design process.The density based and high dimensional space based anomaly detection methods are used respectively for low dimensional detection features and rich model features.Experimental results show that compared with the typical spatial steganography algorithms,the proposed algorithm has better detection efficiency.
-
Computer Engineering. 2017, 43(11): 245-251. https://doi.org/10.3969/j.issn.1000-3428.2017.11.040In order to improve the accuracy of hyperspectral image classification,combined with spectral information,neighborhood information and boundary information,this paper proposes a hyperspectral image classification scheme.The method takes the Local Fisher Discriminant Analysis(LFDA) algorithm to reduce the dimension and get the boundary information.The proposed Block Nearest Classifier(BNC) algorithm is used to get the discriminant information with the spectral feature and neighbor feature.The boundary information is used to smooth the classification label obtained from BNC algorithm.Experiment is carried out on hyperspectral dataset of 3 real ground objects.Results show that the proposed scheme improves the classification accuracy of hyperspectral imag effectively and robust.
-
Computer Engineering. 2017, 43(11): 252-260. https://doi.org/10.3969/j.issn.1000-3428.2017.11.041In the scene of depth image acquisition,in order to use the high-resolution color map of the scene for superresolution upper sampling,this paper proposes an adaptive learning algorithm for local filter kernels using convolutional neural network.It utilizes both the dense/high-resolution color and the sparse/low-resolution depth information to extract the scene informationentirely.Experimental results on the Middlebury and ToFMark datasets show that,compared with traditional depth superresolution algorithms,the proposed algorithm is capable of obtaining the best super-resolution results.Especially in the color and depth edge as well as the texture mismatch region,it has better robustness.
-
Computer Engineering. 2017, 43(11): 261-271. https://doi.org/10.3969/j.issn.1000-3428.2017.11.042For the problems of premature phenomenon and slow convergence rate appeared in Artificial Bee Colony(ABC) algorithm,an ABC algorithm based on the backtracking search is proposed.The Backtracking Search Algorithm(BSA)is used to select and update the colony.Through the random mutation strategy and non-uniform crossover strategy,the new algorithm can enhance the population diversity of the colony algorithm,enables the algorithm jump out of the local optima and has a better global convergence rate.The improved algorithm is used for image contrast enhancement.By searching the optimal parameters α,β of incomplete Beta function,the gray-scale transformation curve is determined.The image gray level is adjusted to improve the image contrast.Simulation results show that the proposed algorithm,has a higher accuracy and faster convergence rate.Compared with the Histogram Equalization(HE) algorithm,the contrast of the image is enhanced effectively.
-
Computer Engineering. 2017, 43(11): 272-276,280. https://doi.org/10.3969/j.issn.1000-3428.2017.11.043The nonlinear illumination changes of local region frequently have a negative impact on the stability of local feature descriptor.In order to solve the problem,a local feature descriptor based on Stationary Wavelet Transform(SWT) and intensity order is proposed.The affine covariant region is detected by using Hessian-Affine detector.The detected region is decomposed by SWT to get multiple low-frequency subbands of different scales as support regions,and the used multiple support regions can effectively reduce the negative impact which is caused by image distortion.The support regions are divided by intensity orders and this ensures that the proposed descriptor is invariant to the monotonous illumination changes.The local feature descriptor is obtained under the local rotation invariant coordinate system.Experimental results show the robustness of the proposed descriptor for the image with viewpoint changes,linear illumination changes,JPEG compression changes and so on.
-
Computer Engineering. 2017, 43(11): 277-280. https://doi.org/10.3969/j.issn.1000-3428.2017.11.044In order to achieve miniaturization and high performance of the sensor array at the same time,the biomimetic coupled processing on the received signals by small aperture array is proposed to extend the effective aperture of the array.Considering the coupled auditory structure of parasitoid fly as a two-input two-output system,on the basis of its vibration equations,the expressions of the output responses are deduced,and the signal processing model is constructed,while the parameters are optimized.The array signals are processed by biomimetic coupling.Experimental results show that the biomimetic coupled processing can amplify the differences between signals of antenna elements,and extend the aperture of array validly,which can achieve array miniaturization on the premise of ensuring high performance.
-
Computer Engineering. 2017, 43(11): 281-285,291. https://doi.org/10.3969/j.issn.1000-3428.2017.11.045Aiming at the application feature of the stereovision measurement in wind tunnel test,a trinocular stereovision measurement system based on linear CCD is designed.The cylindrical lens system based on biotar lens is fixed on the focal plane of CCD,which is used to convert the point light to linear light,and to reduce the lens distortion simultaneously.Adjusted by three-DOF mechanism to CCD,measurement target can form image nearby the focal plane.The virtual calibration target,constructed with coordinate measuring machines,is used to compute the transformation parameter accurately.Example application results show that the system can track the trajectory of the three-degree-of-freedom virtual flight system,and measures the mechanical dump of free-oscillation system accurately.
-
Computer Engineering. 2017, 43(11): 286-291. https://doi.org/10.3969/j.issn.1000-3428.2017.11.046A data acquisition system design scheme based on 6LoWPAN wireless sensor network technology is proposed to meet the basic needs of data resources in the smart city construction process.Based on the C/S architecture,the data server can obtain data from the bottom network,which improves the flexibility and increases the effective data flow ratio at the same time.By designing the embedded gateway based on the network layer with 6to4 tunnel and Netfilter frame,the integration of 6LoWPAN network and the Ethernet is realized.Meanwhile,the adaptability of the gateway is improved,the average forwarding delay of small packets and the packet loss rate can be decreased.C/S architecture is applied between the data consumption terminal and the data server,this can not only ensure the independent design of data consumption terminal according to different application objects,but also reduce the complexity of the data server.Test results show that the proposed system is feasible,convenient to network,easy to maintain and upgrade.
-
Computer Engineering. 2017, 43(11): 292-296,302. https://doi.org/10.3969/j.issn.1000-3428.2017.11.047Aiming at the growing requirement of processing rate of encryption module in the realm of information security,this paper proposes a design scheme of Universal Serial Bus(USB) encryption card based on self-developed security chip.Firstly,a high-performance Chip Operation System (COS) of multilayered structure is developed which configures the USB encryption card to massive storage device,and Bulk-Only protocol is used to communicate with host without installing driver.Meanwhile,extended SCSI private commands are used to complete encryption and decryption operation,which improve convenience and security.Otherwise,a USB high-speed cryptographic algorithm subsystem is proposed which uses Flow Control Unit(FCU) as cooperator.In addition,some accelerating strategies are adopted,such as USB-DMA,software pipeline,AHB-Lite bus,etc.Test results show that the USB encryption card designed by the proposed scheme can realize high-speed SM1 operation,which is convenient to use.
-
Computer Engineering. 2017, 43(11): 297-302. https://doi.org/10.3969/j.issn.1000-3428.2017.11.048In order to research the road network macroscopic automatic condition identification model which is used in traffic guidance,according to the ensemble learning theory,this paper proposes an automatic condition identification method based on strong space integration.It uses K-nearest neighbor rule to find a set of training samples similar to the traffic data to be discriminated,the neighborhood of the data to be discriminated,and excavates the strong space of the automatic condition learner,then outputs the automatic condition level label.Automatic condition confusion matrix,recall rate and precision rate are used in the experiment.Experimental results show that this model can accurately identify the automatic condition of the road network and meet the practical application of the automatic condition identification.
-
Computer Engineering. 2017, 43(11): 303-309. https://doi.org/10.3969/j.issn.1000-3428.2017.11.049High voltage switchgear temperature online monitoring system currently exists mostly presence deficiencies of complexity for installation and maintenance,unreliable sensor power supply and private interface. In view of this problem, this paper designs a new temperature monitoring system based on Object Linking and Embedding for Process Control(OPC) technology which achieves online monitoring and centralized display switchgear temperature information by using Surface Acoustic Wave(SAW)temperature sensor and ZigBee wireless network to transfer temperature information.And software architecture of OPC serveris used which shields private protocol at the bottom of the device to facilitate the realization of seamlessly integrated with other measurement and control platform. Experimental results show that the designed system can overcome the disadvantages of private protocol and not open interface,and has good engineering application value.
-
Computer Engineering. 2017, 43(11): 310-316. https://doi.org/10.3969/j.issn.1000-3428.2017.11.050In order to solve the phenomenon that Simultaneous Localization And Mapping(SLAM) method based on Rao-Blackwellized particle filtering might lead to particle degeneracy and particle depletion,a SLAM optimization method is proposed.To mitigate particle degeneracy,a kind of region Particle Swarm Optimization(PSO) method is introduced to adjust the particles’ proposal distribution.All particles are clustered into several regions and the weighted central position of each region is calculated.With the particle swarm optimization operation,the particles of each region are derived to the regional central position.During the resampling process,a partial Gaussian resampling algorithm is proposed,in which only the particles whose weight is too high or too low will be processed.Experimental results prove that compared with MT-Gmapping method,the improved method can use fewer particles to generate a high-precision map and satisfy the actual requirement.
-
Computer Engineering. 2017, 43(11): 317-321. https://doi.org/10.3969/j.issn.1000-3428.2017.11.051The traditional forecast design method depends on the experiences from the designer,which cannot consider the nature of the wind speed signal changes,it results in the low generality ability of the model structure.Therefore,the RBF neural network in combination with the manifold algorithm is proposed to design the model structure and extract essential features in order to increase the stability and robustness,and improve the forecast accuracy and generality ability.Experimental results using the data from a real wind farm in East China show that,compared with the traditional wind speed forecast methods,the proposed model structure selection method can improve the computing efficiency,reduce sample complexity,and has better forecast effect.
2024
Volume 50 No 4
Volume 50 No 4
(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