2011 Vol. 26, No. 3
Display Method: |
A KSVM classification algorithm by combining the advantages of SVM algorithm and KNN algorithm in machine learning is proposed.Some with semantics tendency and combined the parts of speech is chosen as the characteristic items,and the proposed algorithm is applied in sentiment classification of network comments to judge one comment positive or negative.Experimental results showed that the proposed algorithm can classify some comments online with a higher accuracy.
A new stereo matching approach based on reconstruction error information was proposed.The method reconstructs images using the orthogonal property of Tchebichef discrete orthogonal moments in image coordinate space.The reconstruction error image is computed.Several stereo image pairs are used to compute the disparity images in error image space using stereo match approach based on correlation window.The result indicated that the stereo match approach based on error images is possible and valid though its matching result is not better than the result computed in image space or reconstruction image space.
An outlier mining algorithm based on the clustering ensemble was presented in order to reduce the reliance for users and decrease the high false positive rate due to taking the small size clusters as the outliers directly.Outliers can be found according to the abnormal frequency of every record.The algorithm is able to provide the user a more friendly operation.The experimental results on the real-life datasets showed that the proposed algorithms are feasible and effective comparing with other classical algorithms and can be used for mixed dataset.
For itemset search space is too huge which increases concise association rule based on closed itemset calculation and decreases mining efficiency,a new search space pruning method was proposed.It exploits suffix-tree like structure and just keeps as little as tail information.It can prune search space with little cost and mine concise itemsets and rules directly.Additionally,it avoids scanning original database recursively and achieves good performance.Experimental results show it is an effective method for concise association rule mining and has good scalibility.
A digital watermarking algorithm based on double encryption and discrete Fourier transform(DFT) was proposed.To ensure security,the chaotic scrambling based on logical mapping is conducted before embedding carrier image,then the implementation of encryption by Arnold transform is furthered.Then according to the bit values of watermarking,it is to be embedded to modify the phase of DFT coefficients.Experimental results show that the algorithm has good invisibility and robustness.
The audio watermarking scheme of LSB was improved aiming at its easy operation and its lower robustness.The improved LSB watermarking scheme heightens the LSB layers and ensures the limpidity through the change of the contiguous codes and noise shaping.From the wrong bit rate reports achieved by emulation software,the new LSB watermarking scheme has the better robustness.
Based on the engineering drawing system design and implementation,an engineering drawing retrieval method based on inexact graph matching was presented.A representation of primitive attributed relational graph (ARG) for each engineering drawing is extracted;and the vertex compatibility matrix and edge compatibility matrix between the ARGs of the target and searched model are calculated,the measure of the similarity between the two models is created.The Sinkhorn's alterative normalization method is taken to solve the matching optimal problem.In the matching process,using the ARG's node and edge attributes to pruning the search space.Experimental results show that this method is able to support the inexact model retrieval and its efficiency meets the requirements of practical applications.
The palmprint image is preprocessed and divided the region of interesting (ROI) by the geometric profile of a palm.Then,basing on different wavelet basis functions,the ROI is transformed by the multilayer wavelet transformation respectively,and these different wavelet energy features are extracted later.The recognition ability of different wavelet basis functions is obtained by calculating the recognition rate under different wavelet basis functions and different levels of transformation.Comparing with previous algorithms,the experimental results show that this method has better recognition effect.
To avoid the effect of transmission and acquisition process by mixed noise on visual effect,wavelet packet decomposition and reconstruction algorithms,spatial filter for denoising and wavelet image fusion method were applied.Experiments showed that the wavelet packet transform with a variety of noise can enhance the image processing and remove image noise while preserving details of the background of a clear image enhancement.
A moving object tracking algorithm based on the color information and shape information was proposed,for the humanoid soccer shows poor performance on detecting moving object effect and is easily affected by the variable illumination.It implements rapid image threshold segmentation based on color information in the HSI color space,obtains the object information,and was joined by the adaptive threshold update to improve the robustness of algorithm.The possible position of the moving object in the next frame is predicted by Kalman filter,then,according to the shape information,implements optimizing edge detection to recognize the object in local area,obtains the accurate object location information in the image,then continues to track.The experiments show that this algorithm can accurately track the moving object,satisfy the requirement of the real-time.
Based on the image processing,a new real-time background image update algorithm was proposed,and a fast and robust automatic vehicle identification system was designed.Experimental results showed the feasibility of the program.The α value of the new algorithm is very important for background image update.
A novel algorithm based on a structure space was presented to realize any automatic network model graph generation for complex systems,in order to solve the problem that the distribution network model of complex electromechanical system automatically generates graphs.A structure space is designed as a special data structure,which is used to save the basic information of the network model graph.Then different sets are defined to manage the classified information in the structure space according to certain operating sequence.The drawn model graph is based on classification sets,which contains the computing and the drawing of graphic objects.In the drawing process,the iterative calculation is used,and the whole graph is displayed with the nodes,relationships,and parameters eventually.Finally,a network model is used to draw an example illustrate the proposed algorithm.It shows that the proposed algorithm is feasible and may satisfy the realistic requirements.Simultaneously,it is discovered that the auto-generation arithmetic for the network model graph has a good universality,and can be widely extended in the application.
According to portable communication terminal has limited computation resource and communication ability.Which is difficult to implement key cryptography algorithm with large amount of calculation question.It is proposed an authentication method based on EAP-SAKE in order to solve the problem.It uses some security technologies to provide mutual authentication between peers,such as message integrity code,etc.This authentication method generates different keys for various purposes from root secret and constitutes key hierarchy.In addition,it can resist replay attack and man-in-middle attack.It has less communication load and computing amount and provides key negotiation.So it is more suitable for protecting the information security in mobile terminal which has limited resource in WLAN.
Since the Diffie-Hellman key exchange protocol lacks the authentication for identity,it might be suffered from the man-in-the-middle and replay attack.In order to resolve this bug,a modified Diffie-Hellman key exchange protocol-AADH protocol was put forward.The AADH protocol has introduced some new technologies,such as key authentication,timestamps and random numbers which is the marker of communication.The analysis of the performance showed that due to the safe performance of Diffie-Hellman protocol,AADH protocol has high ability in fighting back the man-in-the-middle and replay attacks.
The Overcast working process and principles of Cisco based on the PeerCast model were expounded,and the node selection strategy of Overcast system was analyzed,the delay performance of bandwidth priority algorithm was improved and node structure tree was also balanced.The simulation results showed that the performance of this new algorithm was better than the original node selection algorithm.
In order to construct the multi-machine experimental platform for networks architecture courses,a virtual machine experimental platform in single PC was proposed by making use of Virtual PC and VMware technology properly.Virtual PC is used to build the single-LAN experimental environment,such as DNS services,Web services,FTP services,email services and streaming media services,et al.VMware is employed to build the multi-LAN experimental environment,such as IP routing,NAT,remote access,DHCP and other experiments.Practice result showed that the proposed virtual machine experiment platform could not only greatly facilitate the students to independently complete the experiments,but also effectively reduce the laboratory operation and maintenance costs.
In order to protect the private network against attacks from both outside and inside efficiently,a new self-defending system based on IPv6 was designed.The system analyses the whole session's data,makes network anomaly detection with MWM pattern-matching algorithm,then makes the content control against disclosing the secrets through the TCP interceptor technology,and realizes self-defending.
For the inadequacies of low power adaptive clustering routing protocols,a new modified forced cluster-head selection approach based on the virtual round was presented.A virtual round only responsible for selecting cluster of is started to help cluster-head selection when the random number generated by all the nodes are not ideal.The simulation on OPNET platform demonstrated that the modified routing protocol enables the whole network have a fine performance.
In order to solve the issues of forwarding packets abnormally,this paper designs the Hub-Spoke topology of the frame-relay multipoint interface network,and configures the non-broadcast multi-point interface network by neighbors discovery,Designated Router/Backup Designated Router election,establishing frame-relay mappings,etc.According to the test,the network operation is stable and reliable,vast engineering personnel with beneficial references in commissioning frame relay multipoint interface network was provided.
Based on the principle of network traffic capacity statistics of website cluster or multi-website,network traffic capacity statistics of website cluster or multi-website analysis system was designed:by means of three-layer design patterns (entity,data accessing,service logic),a layered architecture of the system is presented which realizes traffic capacity statistics of website cluster or multi-website and improves distinctly the performance of the system in load capacity,efficiency and stability.The system can provide a reference for revising or optimizing the strategy of network marketing for user and administration authority.
In order to eliminate noise preferably and extract topic content from Web pages efficiently,an algorithm of Web page purification is presented.This algorithm argues that topic content of Web page is mainly contained in table and p,hereby Web noise can be preprocessed.Then with the content match of relevant Web page,the topic content of Web page can be acquired by way of calculating the importance of node.This algorithm has achieved very precise results,correctly extracting 98.2% of the pages in a set of 6 318 pages in portal sites.When used for data mining tools,this algorithm is better than the other similar algorithms.It can eliminate noise efficiently.
Focused on the lower matching scores of input and output port in every time slot based on input-queuing the maximal matching of router,the proposal is that by changing the methods to update polling pointer so as to improve the number of port matching,solve the synchronization problems on output port,and effectively optimize throughput rate of system.The simulation experiment showed that using matching of polling pointer was able to effectively optimize the performance of switches.
Extensible hashing and linear hashing were discussed,and an improved algorithm for the hashing suffix length inequality was stated,which avoided unnecessary overflow bucket.The number of hash buckets grow linearly,which avoid splitting buckets and updating bucket address table continually,caused by unusual distribution of search key.The experiments of the simulation method showed that the improved algorithm was significantly better than extensible hashing and linear hashing.
To achieve the safety management and early-warning of cold chain,the cold chain food safety control system was set up based on the internet of things.The system collects,transports and disposels the cold chain food information automatically through the core technology of the internet of things,such as the sensor,RFID and GPS,to realize the safety control and pre-warning management.For the system promotion still has some problem,the standard of the things of internet should be established quickly,and much should be put into the key technology to form the core technology with independent intellectual property.Great importance should be attached to the corresponding security issues and actively guide the demonstration and application of the industry,to promote the actual development of the cold chain food safety control system based on the internet of things.
In order to solve the problems about the use of COM components controls in Java development and operating Word document format in the Java program,a JACOB-based Java general WordWriter component was designed using CBSD idea and component recombination technology.This component can easily generate various word documents formats.In addition,it makes the program more readable,easy to maintain and shorten the software development cycle and reduce development costs.
Based on modular design,the remote sensing data record keeping software was designed.The data recording module collects,processes and real-time records the demodulator high-speed output stream of data;FRED format generation module uses the general FRED format to arrange and process remote sensing data,and the remote sensing data archive format can be realized;data playback module playbacks remote sensing data after completion of archives for self-treatment.The results showed the software was feasible and achieved the project target.
A realization method of two-machine serial communication based on LabVIEW was presented,using serial line to design the communication between two computers,and a communications system which can automatically send and receive had been realized through LabVIEW's VISA library.The serial communication control module was built up in a logical sequence,at the same time,looping statements,condition judgment statements and other statements were used and the data transfer format or instrument control command is converted to LabVIEW platform.The method is practical with a simple hardware interface,convenient software programming and practice.It has some reference value in the actual close communication.
An improved distributed denial of service (DDoS) attack detection system was proposed for defending DDoS attack.The system detection was improved using bayesian methods to compute posterior probability and according to the state of the first connection was used.In order to improve system response function,the passive response is used against the intrusions by the real-time response.The result of simulation indicated that the improved system has reduced the rate of false positives and omission,and improved the real-time responsiveness.
An optimization mathematical model of UTP (university timetabling problem) was established based on UTP analysis detailedly,which includes:timetabling target,main influencing factors,constraints,and solving target.The framework structure to solve UTP was founded.A variety of improved GA schemes were designed,which include:decimal encoding scheme,initial population design scheme,fitness function design scheme,niche strategy,adaptive crossover probability and adaptive mutation probability design scheme.Simulation results showed that the proposed GA can satisfy multiple constraint conditions and resolve UTP more effectively.
The MVC design pattern,B/S mode,Ajax,and MD5 encryption technology were used in moral education credit management system platform to get a management information system with good performance.This system has the following functions:activity management,student management,league branch management and teacher management.The test shows this system is scalable,portable,reusable,and secure.
Founded in 1986, bimonthly
Administered by:The Education Department Henan Province
Sponsored by:Zhengzhou University of Light Industry
Editor-in-chief:Wei Shizhong
Executive Editor-in-Chief:Zou Lin
Deputy Editor-in-Chief:Qu Shuanghong
Edited & published by:Editorial Department of Journal of Light Industry
CN 41-1437/TS
ISSN 2096-1553
Address:136 Science Avenue, Zhengzhou City, Henan Province, China
Postal Code:450001
Tel:(086)0371-86608635
(086)0371-86608633