15 September 2018, Volume 44 Issue 9
-
Select all|
-
Computer Engineering. 2018, 44(9): 1-8. https://doi.org/10.19678/j.issn.1000-3428.0048019
Rigid formation of multi-robot can not flexibly adjust formation,leading to the phenomenon of chasing blind angle easily,so that the pursuers can not accurately and efficiently complete the chase.To solve this problem,an adaptive rigid structural formation algorithm is proposed.The formation center controller makes the formation center infinitely close to the target.The formation controller is designed to dynamically adjust the formation according to the location of the target.On this basis,the tracker,in accordance with the position of the target and the environmental conditions,combined with the improved rigid structure method to choose the appropriate formation adaptively,so as to complete hunting.Simulation results show that,the tracking time and energy consumption ratio of the algorithm are lower than that of the rigid formation algorithm,and the hunting blind angle can be avoided.
-
Computer Engineering. 2018, 44(9): 9-14. https://doi.org/10.19678/j.issn.1000-3428.0048175
Aiming at the problem that the traditional micromouse maze search algorithm can not adapt to the random mazemap search,a new micromouse walking maze fusion algorithm is proposed.The algorithm uses the probability distance to divide the maze into eight regions,calibrates the probability distance features of each region and fills the algorithm to achieve efficient fusion of the probabilistic distance centripetal algorithm and the flood algorithm,improving the maze search efficiency and reducing the dependence on the high-performance CPU.Test results of 6 mazes show that,compared with the traditional centripetal and flooding algorithms,the algorithm can reduce the maze search time by 50% and the search success rate by 100%,it is an efficient maze fusion search algorithm.
-
Computer Engineering. 2018, 44(9): 15-21,27. https://doi.org/10.19678/j.issn.1000-3428.0048325
The front pose estimation and back-end optimization of indoor mobile robot Simultaneous Localization and Mapping (SLAM) are susceptible to motion blur.To solve this problem,an indoor location and map building algorithm based on Inliers tracking statistics is proposed.After extracting and matching the features of RGB images and using the RANSAC algorithm to get Inliers,the fuzzy images which are affected by the motion of the camera are eliminated by tracking and statistics of the number of Inliers,and then the position of the camera is solved by the nonlinear optimization method of the Iterative Closest Point (ICP).On this basis,the trajectory and 3D dense point cloud images are obtained through closed loop detection and optimized global pose.Experimental results show that,compared with RGB-D SLAM algorithm,the proposed algorithm can effectively improve the robustness and accuracy of SLAM system map building.
-
Computer Engineering. 2018, 44(9): 22-27. https://doi.org/10.19678/j.issn.1000-3428.0047977
Aiming at the problem that Gmapping algorithm can not be accurately located due to particle depletion in high similarity and multiple closed-loop environment,a Rao-Blackwellized Particle Filtering(RBPF) Simultaneous Localization and Mapping(SLAM) optimization algorithm based on firefly algorithm is proposed.The firefly algorithm is used to improve the estimation ability of the particle filter,and the sampled particle set is moved to the high likelihood region to improve the particle distribution while ensuring the low likelihood particle diversity to reduce the effect of particle deficiency.Simulation results under MIT and FHW datasets show that the optimization algorithm establishes more accurate raster maps in different experimental environments,and verifies its validity and feasibility.
-
Computer Engineering. 2018, 44(9): 28-32,37. https://doi.org/10.19678/j.issn.1000-3428.0050408
The manual inspection methods of traditional substations are subject to many objective constraints,and there are certain problems such as false detection and missing detection.Therefore,a robot inspection method is adopted in real-time positioning of the substation,and an adaptive Monte Carlo algorithm is proposed.Using the open source robot operating system,real-time positioning is realized by combining the laser sensor and the odometer.Simulation experimental in Matlab environment results show that compared with Monte Carlo algorithm,the adaptive Monte Carlo algorithm is more accurate and more anti-interference,and it is suitable for real-time positioning of substation inspection robots.
-
Computer Engineering. 2018, 44(9): 33-37. https://doi.org/10.19678/j.issn.1000-3428.0048260
The traditional hand-eye relationship homogeneous matrix equation solving process is cumbersome and complicated,and the accuracy is susceptible to environmental influences.Therefore,based on the idea of converting image coordinates directly into robot coordinates,using the analysis computing ability of OpenCV library function,a hand-eye relationship calibration method based on two-plane homography matrix is proposed.Using the virtual “nine squares” and a fixed reference point calibration mode,a linear camera imaging model is established,and a two-plane homography matrix is obtained,thereby guiding the manipulator positioning.Experimental results show that the method is only 38 s,and the accuracy is higher than the traditional calibration method.
-
Computer Engineering. 2018, 44(9): 38-44,58. https://doi.org/10.19678/j.issn.1000-3428.0048551
Aiming at the problems of low recognition rate,slow speed,easy to be affected by illumination and weeds effect in the visual path recognition of agricultural robots,a visual navigation path recognition method based on mixed threshold and offset line midline is proposed.The illumination independence is used to eliminate the illumination interference.The mixed threshold method is used to segment the image,and the crop line based on the offset line midline method is given.Experimental results show that compared with the traditional agricultural robot crop line identification method,the method has a small amount of data calculation,the average time is about 200 ms,and the crop line identification accuracy rate is 98%,which can effectively determine the navigation path of the agricultural robot.
-
Computer Engineering. 2018, 44(9): 45-51. https://doi.org/10.19678/j.issn.1000-3428.0050766
Aiming at the problems that the existing cross-social network user identity matching algorithm has low accuracy and difficult data acquisition,a new cross-social network user identity matching algorithm is proposed.Using the known matching account nodes,the network fusion algorithm is used to transform the cross-network problem into a single network problem,and the user name information is vectorized and integrated with the topology information vector,and the network representation learning technology is used to obtain the fusion user name and topology.The account node vector of the two types of information is structured to implement user identity matching.Experimental results show that the average F1 value of the algorithm is 79.7%,which is 7.3%~28.8% higher than the traditional machine learning algorithm and the existing two benchmark algorithms,it effectively improves the user identity matching effect.
-
Computer Engineering. 2018, 44(9): 52-58. https://doi.org/10.19678/j.issn.1000-3428.0049799
Aiming at the problem of spatio-temporal data similarity evaluation with fixed spatial position and trend of attribute value trend,based on the Biharmonic spline to establish the trend surface,a new spatio-temporal data similarity algorithm is proposed.The trend surface image is formed by mesh extraction and color gradation mapping,and the trend state of the spatio-temporal data is characterized as the structural information of the image.The similarity of spatio-temporal data is characterized by the similarity between the trend surface images,and through Structural Similarity(SSIM),the similarity evaluation scheme of spatio-temporal data structure is given,and the similarity synthesis of time dimension is implemented to avoid the one-sidedness of traditional relying on vector space analysis,and a solution for spatio-temporal data similarity analysis is provided under a certain time window.Experimental results show that the algorithm can effectively describe the trend information contained in spatio-temporal data and improve the applicability of such spatio-temporal data similarity algorithm.
-
Computer Engineering. 2018, 44(9): 59-63,69. https://doi.org/10.19678/j.issn.1000-3428.0047866
Point of Interest(POI) recommendation helps users to find the desired location,but the recommendation accuracy of existing recommendation algorithms is low.To solve this problem,a POI recommendation algorithm fusing with spatiotemporal and popularity features is proposed.The time characteristics are integrated into the user based collaborative filtering algorithm,and the estimation of the POI popularity based on time factors is integrated into the spatial characteristics,the corresponding estimation methods are given respectively for the two methods,and a joint framework based POI recommendation algorithm is obtained.Experimental results show that,compared with U,UTF and U+SB algorithm,the proposed recommendation algorithm can effectively improve the recommendation accuracy and recall rate.
-
Computer Engineering. 2018, 44(9): 64-69. https://doi.org/10.19678/j.issn.1000-3428.0048497
Most of the researches of recommendation system are based on single data source,single recommendation algorithm or simple additive fusion,ignoring the importance of data source and dynamic fusion of algorithms,resulting in low recommendation accuracy.To solve this problem,a new dynamic adaptive recommendation algorithm is proposed.The basic data is used to find the artist popularity and time decay factor,which is used as the recommendation data source to reduce the recommendation error caused by the single data source.A more accurate combination model is constructed by weight factor integration based on neighborhood method and matrix decomposition technology.The data source is applied to the model,and the weight factor of the fixed step is used to adjust the proportion of the two methods in the model,and the dynamic adaptive adjustment is realized according to the recall rate of the recommendation result.Experimental results on the real data set show that this algorithm has higher recommendation accuracy than simple additive fusion and FSWA algorithm.
-
Computer Engineering. 2018, 44(9): 70-77. https://doi.org/10.19678/j.issn.1000-3428.0048137
A Similarity of social feature similarity-aware data forwarding algorithm for Device-to-Device(D2D) communication under cellular networks is proposed to improve the efficiency of data forwarding.Firstly,the similarity of user’s social feature is evaluated with user’s multiple social feature and user’s movement track.Then,an overlapping community with multiple user’s social feature is constructed by the weight of similarity of user’s social feature.Finally,it chooses the relay to forward data in same community through local activity of inner community and forwards data in different communities through the attributes of overlapping and similarity.Experimental results show that,the data forwarding algorithm proposed by this paper can improve the efficiency of data forwarding by the help of user’s attributes of social feature.Compared with the algorithms of data forwarding without considering of social feature,it can improve the average delivery ratio by 36.2% and reduce the average delay of data forwarding by 32.8% when the number of user is 78.
-
Computer Engineering. 2018, 44(9): 78-82. https://doi.org/10.19678/j.issn.1000-3428.0048319
In spatial spectrum estimation,most of the algorithms based on feature subspace require the number of sources to be accurately known.When the number of sources is estimated to be error,the performance of the algorithm is seriously invalid.For the existence of coherent sources,the information theory criterion estimates the number of sources is not accurate,and the linear array-based coherent source estimation algorithm can not be directly applied to the circular array.A circular array coherent source number estimation algorithm is proposed.Through the virtual translation of the array,the complete decoherence of the source is realized,and the information theory criterion is used to estimate the number of sources after decoherence.The mode space transformation of the circular array to the linear array is not needed,and the algorithm complexity is significantly reduced.Experimental results show that the proposed algorithm can correctly estimate the number of coherent sources without loss of array aperture,and the performance is better than FBSS_MDL and Toep_MDL algorithms at low SNR and small snapshot.
-
Computer Engineering. 2018, 44(9): 83-88. https://doi.org/10.19678/j.issn.1000-3428.0047897
The degree distribution function is the key factor affecting the performance of digital fountain codes.The Robust Soliton Distribution(RSD) has the problems of high complexity of encoding and decoding,low decoding performance when the number of original data packets is small in LT cods.To solve this problem,a new truncated robust soliton distribution optimization algorithm is proposed.After adjusting the Stationary Distribution(SD),the RSD is truncated,and the truncated RSD and the adjusted SD phase are merged to be normalized to get a new degree distribution.Simulation results show that,compared with the RSD and SD degree distribution,the proposed algorithm can reduce the decoding overhead of 25%~31% and save the average encoding and decoding time of 22%~55%,and the smaller the number of original data groups,the more obvious the performance improvement.
-
Computer Engineering. 2018, 44(9): 89-94,100. https://doi.org/10.19678/j.issn.1000-3428.0048271
On the basis of the classical Optimized Link State Routing(OLSR) algorithm,a routing algorithm with low delay and high reliability based on absorption mechanism and load balancing mechanism is proposed to solve the problems of low reliability,poor timeliness and dynamic change of traffic load in the aeronautic swarm network.Through the absorption mechanism,when multiple Multi-Point Relay(MPR) candidate nodes have equal opportunity to become MPR nodes,they choose nodes with higher absorbency to reduce the number of flood TC packets in the network.The load balancing mechanism is designed and predict the next time node load state through the ARIMA-SVR combination prediction model to avoid network congestion.Simulation results show that,compared with the classical OLSR algorithm,the proposed algorithm has less TC packet number,lower delay and higher packet delivery rate,and can be better adapted to the aeronautic swarm network.
-
Computer Engineering. 2018, 44(9): 95-100. https://doi.org/10.19678/j.issn.1000-3428.0047728
To solve the problem of shortage of spectrum resources,a power control algorithm is proposed under the condition that the Quality of Service(QoS) of Macro-cellular User Equipment(MUE) is satisfied.It approximates the nonconvex objective function in three different approaches,and converts the problem into a series of convex subproblems,then solves the subproblems iteratively and the solution of the original nonconvex objective function is obtained finally.Simulation results show that after several iterations,the proposed algorithm can effectively obtain the solution of the original optimization problem and improve the system throughput.
-
Computer Engineering. 2018, 44(9): 101-106,112. https://doi.org/10.19678/j.issn.1000-3428.0047886
Using IEEE 802.15.4 devices need an open network layer standard to achieve interconnection.By analyzing the characteristics of IEEE 802.15.4 and IPv6 protocol,IPv6 is selected as the IEEE 802.15.4 network layer standard,and the architecture of wireless sensor network based on 6LoWPAN is established.Based on this,the construction mechanism of 6LoWPAN Wireless Sensor Network(6LoWPAN WSN)based on beacon frame is proposed.The construction mechanism includes address configuration,intra-network routing algorithm and inter-network routing algorithm.When routing between networks,the nodes send data from the application layer to the associated nodes,and associated nodes can simplify the data packets transmitted by piecewise reorganization of the adaptation layer,and transmit the data packets in the network in the way of data frames,unencapsulate when they arrive at the destination node,and restore the required data packets.Simulation results show that compared with LOAD mechanism,the mechanism can reduce the network construction overhead,shorten the delay of routing and reduce the energy consumption.
-
Computer Engineering. 2018, 44(9): 107-112. https://doi.org/10.19678/j.issn.1000-3428.0048076
The existing Virtual Network(VN) mapping algorithms have poor correlation between node mapping and link mapping,and the longer link mapping path leads to lower request acceptance rate and lower revenue overhead.To solve this problem,a VN mapping algorithm based on adjacent node and topology awareness is proposed.Considering node location constraint,node resource degree,closeness centrality and breadth first search algorithm are combined to improve node mapping.At the same time,link map is implemented by link bandwidth requirement and shortest path algorithm.Simulation results show that,compared with G-VNM,TCF-VNM and TAS-VNM algorithm,the proposed algorithm can shorten the proposed link mapping length and improve the request acceptance rate and the return cost ratio.
-
Computer Engineering. 2018, 44(9): 113-117. https://doi.org/10.19678/j.issn.1000-3428.0047433
Underwater Wireless Sensor Network(WSN) has the characteristics of high energy consumption,low efficiency of data transmission,dynamic topology.Aiming at the mobility advantage of nodes in the underwater environment,this paper puts forward a Vector and Energy based Routing(VER) protocol.This protocol builds a vector from the source node to the sink node,then determines the priority according to the distance from the candidate node to the vector.Considering the distance,the remaining energy of the candidate nodes is combined to determine whether it becomes the best next hop node.Simulation results show that compared with VBF protocol and LB-AGR protocol,the algorithm achieves satisfactory results in terms of delivery rate,average end-to-end delay,and energy consumption even if the node’s location is not known.
-
Computer Engineering. 2018, 44(9): 118-123,135. https://doi.org/10.19678/j.issn.1000-3428.0049197
Aiming at the enumeration problem of RapidIO routing network,a flow driven mechanism based on minimum isolation block is proposed.The concept of minimum separation block is given,and the isolation block is divided into the routing network.Then a depth first traversal search is carried out for the entire RapidIO routing network.And the flow driven mechanism is introduced to compare the energy consumption value,power consumption value and the average operation time of the RapidIO switches before and after isolation,to verify the reliability of the algorithm through the flow driven energy algorithm and flow path allocation algorithm.Experimental results show that the enumeration number of RapidIO switches can be reduced and repeated enumeration can be avoided.With the expansion of the routing structure layer,the isolation effect is more obvious.
-
Computer Engineering. 2018, 44(9): 124-129. https://doi.org/10.19678/j.issn.1000-3428.0047780
In the Multi-downlink network,the interference among nonorthogonal channels and the malicious interference makes destination nodes not meet the demand of Quality of Service(QoS).So a security transmission method based on zero-sum game model is proposed to solve this physical layer security problem.Utility function based on secrecy rate is built,and the downlink beamforming and power allocation are optimized jointly based on uplink downlink duality to minimize the transmission power and to meet the QoS.The mixed strategy equilibrium and the corresponding secrecy rate can be obtained.Simulation results show that this method can achieve a good secrecy rate.
-
Computer Engineering. 2018, 44(9): 130-135. https://doi.org/10.19678/j.issn.1000-3428.0048266
A new integrity verification scheme for cloud data is proposed to solve the problems of dynamic data updating and user revocation.The new scheme uses proxy re-signature to transfer data ownership and revoke users.By using the random masking technique,it can prevent the curious third-party auditor from restoring raw data blocks and ensure privacy of cloud data.In addition,a simple authentication structure is constructed by multi-branch path tree,which simplifies the data update process and supports the update operations such as inserting,modifying and deleting the cloud data.The analysis results show that the proposed scheme satisfies existential unforgeability,and can resist replay attack and deletion attack.
-
Computer Engineering. 2018, 44(9): 136-140,148. https://doi.org/10.19678/j.issn.1000-3428.0049312
In order to research the law of rumor spreading and take effective measures to reduce the impact of rumor,according to the principle of social identity,this paper abstracts and quantifies two important behaviors:guiding behavior and imitating behavior.It improves the Bass model,proposes Bass-SI model about rumor diffusion and makes the simulation under Mathematica for the propagation property of the rumor.At the same time,it analyzes the impact of behavioral intensity and behavioral stability on rumor spreading process and proposes the restrain measures.It finds the following conclusion:the guidance behavior determines the early trend of the rumor diffusion process,if the intensity of behavior increases and the stability decreases,the degree of rumor diffusion will increase;the imitating behavior has an important influence on the latter process of rumor diffusion,if the intensity of behavior increases and the stability increases,the degree of rumor diffusion will increase.
-
Computer Engineering. 2018, 44(9): 141-148. https://doi.org/10.19678/j.issn.1000-3428.0049696
In order to solve the problem that the authentication cipher security risk of the traditional 1-to-1 and m-to-1 authentication mode is easier to leak,the paper proposes a m-to-n authentication mode,which includes m clients and n authentication servers.m clients request to n servers at the same time.In this mode,m clients and n servers are abstracted as the secure two-group computation model.It defines the model security under the semi-honest model,puts forward a security multi-party summation protocol.Taking the security multi-party summation protocol as the basic building block and the forgetfulness transfer protocol as the basic cryptographic tool,this paper designs a secure two-group authentication protocol.Analysis results show that this protocol has higher security and efficiency,it can solve the problem of security authentication between intra group and inter group .
-
Computer Engineering. 2018, 44(9): 149-152. https://doi.org/10.19678/j.issn.1000-3428.0048026
Aiming at the problem that the efficiency of the integer Fully Homomorphic Encryption(FHE) scheme is low and the public key is long,Coron’s public key compression technique is used to encryption algorithm with a quadratic form,a partial homomorphic encryption scheme with the property of reducing public key size to O～(λ3.5) is proposed.This scheme can encrypt n bit plaintext once.Compared with DGHV scheme,the proposed scheme has lower public key size and higher encryption efficiency,which is more suitable for the practical application of cloud computing.
-
Computer Engineering. 2018, 44(9): 153-158. https://doi.org/10.19678/j.issn.1000-3428.0047591
In order to effectively solve the classification performance of massive and complex intrusion detection data,an intrusion detection method of hybrid deep learning model is proposed,which is based on Deep Belief Network (DBN) and Extreme Learning Machine(ELM).DBN-ELM uses DBN to extract features from the massive,complex and unlabeled data to get highly abstract features at first,and then takes ELM as the classifier to finish the last classification.It totally improves the recognition rate of intrusion detection and the efficiency of the algorithm operation because it combines the ability of DBN to automatically extract features and fast learning and good generalization of ELM.Compared with DBN,ELM and DBN-ELM,experiments on KDD99 and NSL-KDD dataset show that DBN-ELM has better accuracy and efficiency of algorithm.
-
Computer Engineering. 2018, 44(9): 159-163,170. https://doi.org/10.19678/j.issn.1000-3428.0048993
According to the characteristics of microblog spammers,this paper proposes a semi-supervised spammer identification method based on cost-sensitive learning in latent space.Firstly,it selects twenty-two features of microblog account from perspectives of contents,activities and social relations.Then,it obtains latent account vectors using matrix factorization method and constructs a novel cost-sensitive semi-supervised classification model with the maximum margin theory in latent space.In addition,a social relation regularization from following behaviors is formulated on the model.Finally,it develops a linear-complexity algorithm for solving the model with the stochastic gradient descent method.Experimental results show that the proposed method outperforms existing methods significantly,such as SMFSR and L2-SVMS,in terms of the evaluation measures of accuracy,recall and F1 score.It also obtains nearly linear training speeds.
-
Computer Engineering. 2018, 44(9): 164-170. https://doi.org/10.19678/j.issn.1000-3428.0048518
Entity relation extraction technology aims to determine the relationship among entity pairs of statements via the text content.Due to the complexity of the grammatical structure,the variety of word meanings and other factors in Chinese,the relationship classification of Chinese entities is very limited.Aiming at this problem,this paper presents a deep learning method expressing texts based on the Shortest Dependency Path(SDP),which makes full use of the dependency analysis to express the statement and word class.It uses the bidirectional structure with Long Short-Term Memory(LSTM) units while learning the information expressed by the SDP,and then uses the Convolutional Neural Network(CNN) training classification model for LSTM output.Experimental result shows that this method can extract the entity relation accurately,its F1 value is higher than that of CNN method and BI-LSTM method.
-
Computer Engineering. 2018, 44(9): 171-176. https://doi.org/10.19678/j.issn.1000-3428.0048713
To choose the most appropriate answer from the waiting list as the title of the article,this paper proposes a model which is based on Convolutional Neural Network(CNN) and Recurrent Neural Network(RNN).The model contains the characteristics of extracting local data features from convolutional neural network,combines the characteristics of capturing association information between sentence contexts from recurrent neural network.The method excavates correlation from the article and the title to select the best answer.Experimental results show that this model can effectively shorten the length of text and improve the accuracy of extraction.
-
Computer Engineering. 2018, 44(9): 177-183. https://doi.org/10.19678/j.issn.1000-3428.0051133
Traditional trajectory prediction algorithms require a lot of time when training models,the complexity of time and space is too high,and the execution efficiency is too low to meet the needs of real-time prediction.Therefore,this paper proposes an advanced mobile target trajectory prediction algorithm.The algorithm calculates the trajectory similarity based on Euclidean distance to improve the accuracy and effectiveness of the prediction.The predicted trajectory is simplified based on the principle of minimum description length,which optimizes the calculation efficiency and trend display.Experimental results show that the algorithm can accurately predict the trajectory of the mobile target,and has a lower complexity of the algorithm,which is completely suitable for the trajectory prediction of the mobile target under the massive data background.
-
Immune Bidirectional Frog Leaping Algorithm and Its Application in Multi-modal Function OptimizationComputer Engineering. 2018, 44(9): 184-191. https://doi.org/10.19678/j.issn.1000-3428.0048176
In order to overcome the shortcomings of Shuffled Frog Leaping Algorithm(SFLA) in solving multi-modal function optimization problems which includes easily falling into local minimum,low accuracy and difficultly searching extreme points as many as possible,a novel Bidirectional Immune Shuffled Frog Leaping Algorithm(BISFLA) based on circle derived mutation is proposed.This algorithm in each loop iteration employs the form of local-global based on bidirectional evolution mechanism to searching in the feasible zone,and then uses double Control Mutative Clonal Selection Algorithm(DCSA) to improve the accuracy of the sub-optimal solution.The function is rotated to further verify the performance of the algorithm.Simulation results show that compared with the original frog leaping algorithm,the optimization accuracy of the algorithm and the number of extremum points are significantly improved while the convergence speed is guaranteed.
-
Computer Engineering. 2018, 44(9): 192-198. https://doi.org/10.19678/j.issn.1000-3428.0048352
To improve the problem that random strategies adopted by Fuzzy C-Means(FCM) and other clustering algorithms lead to the inconsistent results of overlapping community detection in complex network when they are repeatedly run,an overlapping community discovery algorithm is proposed.Node importance is introduced to quantify kernel nodes and some kernel nodes can be chosen by the sequence of node importance and the shortest path between nodes to initialize the cluster centers of FCM,which can improve the instability of FCM.The similarity measure based on the shortest path is provided to obtain the similarity matrix which contains much more information and improve the accuracy of the proposed algorithm.The community affiliations of nodes are determined on the ground of threshold after running modified FCM in the space transformed by spectral clustering.Experimental results show that the proposed algorithm can achieve the only community division result,and it leads more than 8% for the NMI in datasets of the Karate and Dolphins compared with GCE,INFOMAP,COPRA and other algorithms.
-
Computer Engineering. 2018, 44(9): 199-202. https://doi.org/10.19678/j.issn.1000-3428.0047961
In order to solve the problem of face image super-resolution when the illumination changes,a robust face super-resolution algorithm for illumination is proposed.Combining the input low-resolution face image with the face image training set,the low resolution face images under multiple illuminations are generated by the diagonal loading redundant transformation in the low resolution space,and the low resolution face images under multiple illuminations is reconstructed by geometry and position constraints.When the optimal weight is solved,the high resolution face images under multiple illuminations are synthesized.Experimental results show that the proposed method cannot only produce the low resolution face images with multiple illuminations for a very low-resolution face images in low resolution space,but generate satisfying high-resolution face images of multiple illuminations.
-
Computer Engineering. 2018, 44(9): 203-211,217. https://doi.org/10.19678/j.issn.1000-3428.0048181
Existing saliency detection method based on graph-based manifold ranking has the problem that the foreground seeds cannot be accurately selected because of idealistic prior background assumption,which influences the detection result.For this problem,a new saliency detection algorithm is proposed.The convex hull is got by calculating Harris feature corner of an original image to roughly estimate the location of salient region.Then,the original image is segmented into superpixels and the foreground seeds can be selected by calculating the dissimilarity between the superpixels inside and outside the convex hull.Two close-loop undirected graphs are constructed by using different regional feature descriptors.Accordingly,taking the selected foreground seeds as query nodes,two saliency maps are generated by adopting classical graph-based manifold ranking algorithm based on corresponding graphs.These two saliency maps are integrated and the result is further optimized to obtain a final saliency map.Experiments on the public MSRA10K and ECSSD data sets are done and the results show that the proposed method can outperforms the traditional graph-based manifold ranking algorithm on better precision rate,recall rate,F-measure and MAE value,and the detection result of the proposed method not only can accurately highlight the salient object in an image,but can suppresses the interference of the background area effectively.
-
Computer Engineering. 2018, 44(9): 212-217. https://doi.org/10.19678/j.issn.1000-3428.0047934
In view of the fact that the image super-resolution reconstruction method cannot effectively reconstruct more high frequency image information in the process of image processing and storage,this paper proposes an image super-resolution reconstruction method based on sparse representation combined with guided filtering.The high frequency image information can be recovered further during image super resolution.At the same time,image matting is introduced in this algorithm,so that the low-resolution image to be reconstructed is decomposed into foreground,background and edge layer by guided filtering.The clarity of the image is mainly determined by the edge layer,so the edge layer of high resolution image can be represented as a sparse linear combination in a dictionary trained from training images,while the foreground color and the background color are reconstructed by bicubic interpolation,then the final high-resolution image is obtained via image synthesis.Experimental results show that the peak signal-to-noise ratio and the structural similarity of the reconstructed image are improved,and the visual effects of the image are also improved,compared with image super-resolution reconstruction via sparse representation.
-
Computer Engineering. 2018, 44(9): 218-223. https://doi.org/10.19678/j.issn.1000-3428.0047769
In order to take advantage of multi-core computing resources to improve the real-time performance of image matching identification,image matching parallel algorithm based on improved minimum distance method is designed.The standard euclidean distance is chosen to replace the simple euclidean distance as the similarity measure of the features,and the MPI+OpenMP hybrid programming model is used to verify the algorithm on multi-core computers.The results show that,the algorithm can reduce the matching error and matching operation time,moreover,MPI+OpenMP hybrid programming model has good speed-up ratio and parallel computational efficiency.
-
Computer Engineering. 2018, 44(9): 224-229. https://doi.org/10.19678/j.issn.1000-3428.0048000
Aiming at the problem of poor performance and high computational complexity in the image fusion algorithm which based on Pulse Coupled Neural Network(PCNN) in the Non-subsampled Contourlet Transform(NSCT) domain,an fusion algorithm based on Compressed Sensing(CS) and adaptive PCNN in NSST domain is proposed.The source image is decomposed into high and low frequencies in the NSST domain.An improved PCNN is used to fuse the low frequency subband coefficients.The sum of the squared differences of the pixels is used as the excitation.The sum of the direction gradients is used as the link strength,and the high frequency which need lots of calculations are processed using CS,and the fused image is obtained by inverse transform of NSST.Experimental results show that the algorithm performs better in information entropy,spatial frequency,edge information evaluation factor,and the running time is shorter compared with NSCT fusion algorithm,NSST combined with PCNN algorithm,and so on.
-
Computer Engineering. 2018, 44(9): 230-235. https://doi.org/10.19678/j.issn.1000-3428.0047951
The current Markov has a problem of invalid iteration compensation for a smooth image block,increasing the time complexity.For this reason,this paper proposes a Spatial Error Concealment(SEC) algorithm.The proposed algorithm utilizes the mean-absolute-distortion criterion to measure the variation of iterative gains,and pre-determines an appropriate threshold to filter out invalid iteration process of the existing algorithm,and thus reduces the average iteration times of whole block losses.Experimental results show that the proposed algorithm significantly reduces the run-time while guaranteeing the similar concealment quality.
-
Computer Engineering. 2018, 44(9): 236-242. https://doi.org/10.19678/j.issn.1000-3428.0048152
To improve the accuracy of the nasal tip point detection and posture correction,a three-dimensional human face nasal tip point detection and posture correction algorithm based on effective energy is proposed,which combines the judgment ability of effective energy for the surface concavity and convexity.Firstly,the nasal tip candidate points are determined by effective energy in the pre-processed 3D face image,then the second screening is carried out according to the mean and variance of the effective energy.The densest region with the most intensive candidate points is selected as the nasal tip region to further obtain the tip of the nose.The effective face area is cut off after calculating the geodesic distance from the tip of the nasal to the other points on the three-dimensional human face,the principal component analysis is used to make correction for the cut face area,and the alignment error is obtained by precise alignment of the Iterative Closest Point (ICP) algorithm.The comparison experiment is simulated based on a large number of 3D face data,and the results show that the nasal tip point detection and posture correction success rate of proposed algorithm respectively reach 98.8% and 98.3%,and the operation time is only 1.489 8 s.
-
Computer Engineering. 2018, 44(9): 243-249. https://doi.org/10.19678/j.issn.1000-3428.0048138
For the traditional Two Dimensional Convolutional Neural Network(2D-CNN),the time dimension information is lost,and thus the dynamic gesture cannot be recognized.This paper proposes a novel dynamic hand gesture recognition method based on video streams,which can effectively improve the overall performance of hand gesture recognition.The input data is simply preprocessed.The spatio temporal feature extraction operation is performed by using improved Multiple Dimensional Convolutional Neural Network(MD-CNN).A multi-sensor fusion method is provided and the dynamic gesture recognition is realized by using Support Vector Machine(SVM).Experimental results show that the proposed method performs well in robustness with respect to the gesture background and illumination.Furthermore,the method achieves the high recognition accuracy beyond 87% for every kind of dynamic gesture dataset.
-
Computer Engineering. 2018, 44(9): 250-255,262. https://doi.org/10.19678/j.issn.1000-3428.0048079
Aiming at the problem that the temporal continuity of face position and the real-time requirement of face detection are high in the current video face replacement research,a video face quantitative detection algorithm based on the interframe constraint model is proposed.The interframe constraint model is established by using the correlations of the frames before and after,and the specific position of the video face is quantitatively described to avoid the deviation of single frame detection.At the same time,adaptively changing the face search area improves the running speed of the algorithm.Due to the similarity of eyebrows and eyes,the false detection and missed detection of face detection algorithms based on Haar_like features is reduced by adding two kinds of Haar_like EB features.Experimental results show that the algorithm can improve the temporal continuity of the quantitative detection of face position in video,and it can improve the running speed of video face detection and reduce the false detection rate.
-
Computer Engineering. 2018, 44(9): 256-262. https://doi.org/10.19678/j.issn.1000-3428.0047850
The existing open-loop mode selection algorithm relies on the accuracy of signal classification,but in most cases the accuracy is low,resulting in poor coding quality in open-loop mode.Therefore an improved neural network based open-loop mode selection algorithm is proposed.The decision tree algorithm of the original open-loop mode selection is replaced by the neural network,and the closed-loop mode selection is selected for training to obtain the mode selection classifier.According to the logic process of the closed-loop mode selection,the neural network is used to predict the input signal,and the two codes are ACELP256 and TVC256.The signal to noise ratio of the mode replaces the signal-to- noise ratio that the coding attempt is calculated.Experimental results show that the accuracy of the two methods is 5.96% and 18.07%,respectively,and the accuracy of music classification is 3.84% and 20.29%,the performance of subjective and objective tone has high improvement.
-
Computer Engineering. 2018, 44(9): 263-268,273. https://doi.org/10.19678/j.issn.1000-3428.0047869
When the target is completely occluded,the algorithm will appear to track the loss of the target.For this reason,an improved spatio-temporal context visual tracking algorithm based on particle filtering is proposed.The Bhattacharyya calculated from the color histograms of the two samples is used to determine whether the target has a global occlusion.When the Bhattacharyya is smaller than the set threshold,the algorithm uses particle filtering to predict the target tracking result.Experimental results show that compared with the original spatio-temporal context algorithm,the algorithm can improve the tracking success rate to more than 85%,the center position error is small,and the robustness is better.
-
Computer Engineering. 2018, 44(9): 269-273. https://doi.org/10.19678/j.issn.1000-3428.0047419
Because of the reverberation in the actual indoor environment,the accuracy of sound source localization is reduced.To solve this problem,a novel localization algorithm based on frequency division and generalized cross-correlation time delay estimation is proposed.The binaural signal gets the subband signal of the frequency division through the Gammatone filter,and performs the minimum phase decomposition in each subband signal,and then calculates each subband after inverse transformation to the time domain from the cepstral domain.Then each subband cross-correlation function is obtained to get the binaural corresponding to the maximum correlation value.The difference between them is finally obtained through geometric models to obtain azimuth angles.Simulation results show that,compared with the PHAT-GCC source localization algorithm,the proposed algorithm can reduce the location error caused by reverberation environment and improve the stability of sound source localization.
-
Computer Engineering. 2018, 44(9): 279. https://doi.org/10.19678/j.issn.1000-3428.0047862
In the process of classifying video texts in the field of science and technology,it is easy to ignore the terminology with high classification contribution.Considering the problem that the traditional Labeled Latent Dirichlet Allocation (LDA) model has biased high frequency words,this paper improves it and establishes the MulCHI-Labeled LDA model for video texts in the scientific field,by building domain termbases to highlight terminology and using chi-square weighting and text position weighting algorithms to improve topic quality.The experimental results show that,compared with the Labeled LDA model,the proposed model can solve the neglect of professional terms and effectively improve the quality of topic words and classification accuracy.
-
Computer Engineering. 2018, 44(9): 280-285. https://doi.org/10.19678/j.issn.1000-3428.0048623
When Time-Triggered Ethernet(TTE) use Time-Triggered Rate Monotonic Scheduling(TT-RMS) algorithm to generate communication schedules,there is a problem that RC messages cannot arrive at the destination terminal within the maximum delay,and TT messages cannot reach the destination terminal after missing the time window.To solve this problem,a scheduling algorithm based on greedy thought is designed.By planning new paths,subpackages and packages,and changing TT message sending time,the communication messages are improved,and a cost function is constructed to evaluate the scheduling table that is generated each time, and the final result is obtained for multiple iterations until the value is stable.Simulation results show that,compared with the improved TT-RMS algorithm,the proposed algorithm can improve the real-time performance of RC messages and reduce the average delay percentage of TT messages.
-
Computer Engineering. 2018, 44(9): 286-295,300. https://doi.org/10.19678/j.issn.1000-3428.0046993
Most of the existing Digital Elevation Model(DEM) multi-resolution representation method are redundant transformation.The redundancy increases sharply with the increase of the scale of the transformation,so it cannot effectively express the multi-scale terrain contour structure.Therefore,a multi-resolution representation method of DEM is proposed,which is non-redundant and can effectively represent the topographic features of different scales.The multi-scale transform structure of non-redundant fractional band wavelet transform is integrated with the multidirectional filter bank of contourlet transform,and the non-redundant fractional decimal contourlet transform is constructed.The experimental results show that,the proposed method can preserve the macroscopic statistical features and local terrain details of DEM effectively,has no redundancy,and the complexity of the algorithm is low.
-
Computer Engineering. 2018, 44(9): 296-300. https://doi.org/10.19678/j.issn.1000-3428.0048184
At present,the demand of large-scale public application to BeiDou high precision navigation and location service is increasing.To solve this problem,the key technology flow of BeiDou high precision positioning data processing and network broadcasting is proposed.Based on the foundation enhancement system of the BeiDou foundation of Sichuan province,the high precision positioning data broadcasting system based on the network mode is designed and realized by the message queuing technology,so as to provide real-time and dynamic high precision data broadcasting service for massive users.Test results show that,the positioning accuracy of the proposed positioning system is less than 0.7 m and the height accuracy is less than 1.5 m.
-
Computer Engineering. 2018, 44(9): 301-308. https://doi.org/10.19678/j.issn.1000-3428.0048036
In the Advanced Orbiting System(AOS) space communication system,more data service types and larger channel fading time-varying lead to low bandwidth utilization,low throughput and unsatisfied QoS service.To solve this problem,a cross- layer optimization method based on utility function maximization of bandwidth and power allocation is proposed.Firstly,the cross-layer utility function model is established,which consider the service quality requirements of the application layer,the Virtual Queue Backlog Ratio(VQBR) of the data link layer and the transmission power and channel state of the physical layer.Secondly,the cross-layer utility function is transformed into nonlinear integer programming problems.Finally,the bandwidth and physical layer transmit power are allocated reasonably by solving the utility function.Simulation results show that,compared with the traditional AOS method and the typical cross-layer resource optimization method,this method can guarantee the fairness of the service,reduce the power consumption,improve the bandwidth utilization and improve the system throughput under the requirements of business QoS.
-
Computer Engineering. 2018, 44(9): 309-313. https://doi.org/10.19678/j.issn.1000-3428.0049602
In order to improve the efficiency of data transmission and the quality of service of users,this paper makes theoretical and empirical analysis in the framework of email systems and message passing mechanism using the queueing theory and network flow.Firstly,a passive measurement is employed,namely traces are collected and email characteristics are mined.Secondly,a basic model is built when the email message delivery system is conforming to the Queueing Theory.Thirdly,the distribution of email sending and the email delivery are calculated and the Kolmogorov-Smirnov(K-S) test show that the statistic gives good over the test of fit to a Poisson distribution.Then,considered the multi- delivery,the distribution of email delivery obeys Erlang distribution and the modified M/Ek/1 model is built.At last,the proposals of improving the efficient of the email transfer are given.
-
Computer Engineering. 2018, 44(9): 314-320. https://doi.org/10.19678/j.issn.1000-3428.0050149
In order to solve the problem that the artificial potential field method for Unmanned Aerial Vehicle (UAV) flight path planning is easy to fall into the local optimal and jitter in the narrow channel,a path planning algorithm based on potential field method improved by collision cone is proposed.Firstly,the determination coefficient of the repulsive potential field of the obstacle is designed by using the collision detection conditions of the collision cone,and the possible adverse effects of the unrelated obstacle potential field are excluded.Then,according to the threat degree of obstacle type and the relative state threat degree of UAV and obstacle,the fuzzy repulsion gain adjustment coefficient is designed to solve the problem that the artificial potential field tends to fall into the local optimal.In addition,in order to prevent the jitter problems caused by changing of potential field and passing through the narrow channel between the obstacles,the jitter adjustment coefficient is also designed and the track point planning method under jitter condition is given.Finally,the obstacle avoidance path planning algorithm for UAV is designed based on the improved artificial potential field function of collision cone.Experimental results show that,compared with the artificial potential field obstacle avoidance algorithm,the proposed algorithm can solve the problem of local optimal and jitter,and it has achieved good results in avoiding dynamic obstacles.
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