Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文

Judgement and Disposal of Academic Misconduct Article Copyright Transfer Agreement Proof of Confidentiality Requirements for Electronic Version More.... Chinese Association of Automation National ResearchCenter for Intelligent Computing System Institute of Intelligent Machines,Chinese Academy of Sciences More....

2011 Vol.24 Issue.3, Published 2011-06-25     Articles 305 Hierarchical Image Automatic Annotation Based on Discriminative and Generative Models KE Xiao, LI Shao-Zi, CAO Dong-Lin Image automatic annotation is a significant and challenging problem in pattern recognition and computer vision. Aiming at the problems that the existing models have low utilization and they are affected by unbalanced positive and negative samples, a hierarchical image annotation model is proposed. In the first layer, discriminative model is used to assign topic annotations to unlabeled images, and then the corresponding relevant image sets are obtained. In the second layer, a keywordsoriented method is proposed to establish links between images and keywords, and then the proposed iterative algorithm is used to expand semantic words and relevant image sets. Finally, a generative model is used to assign detailed annotations to unlabeled images on expanded relevant image sets. Hierarchical model uses less relevant training images to obtain better annotation results. Experimental results on Corel 5K datasets verify the effectiveness of proposed hierarchical image annotation model. 2011 Vol. 24 (3): 305-313 [Abstract] ( 1525 ) [HTML 1KB] [ PDF 653KB] ( 658 ) 314 A Global Scan Matching Method Based on Geometric Statistic Features SUN Feng-Chi, YUAN Jing, HUANG Ya-Lou A global scan matching method is proposed to study how to synchronize robot localization, environmental profile characterization and changed regions discrimination in dynamic environment. In the proposed method, a probeverifysolve (PVS) matching strategy is described. Highlevel geometric statistic features are defined for laser scan segments and scan points. After precise matching between scan data, the environmental profile can be characterized and meanwhile the robot pose can be rectified. Changes in the environment are discriminated using PVS matching strategy. Finally, the proposed method is verified through experiments using several real laser datasets in unknown dynamic indoor environments, and the effectiveness of the proposed method is proved. 2011 Vol. 24 (3): 314-320 [Abstract] ( 860 ) [HTML 1KB] [ PDF 473KB] ( 648 ) 321 Quadric Surface Identification from WireFrame Models ZHAO Jun, GAO Man-Tun, WANG San-Min The angle between two connected curves on same quadric is defined and two minimum internal angle algorithms(MIA_cw and MIA_ccw) that can be applied to quadric surfaces are proposed. When a circle is searched by MIA_cw or MIA_ccw, the vector mixed product is employed to verify whether the circle is a candidate surface or not. In the proposed method, no matter which curved edge of wireframe model is chosen, the candidate surface containing it can be obtained. Thus, it removes the restriction on the start edge in searching candidate surface in existing methods. The candidate surface with curves is separated from those without curves by the proposed method and different algorithms can be used for these two kinds of candidate surfaces. And it requires less time than the one using one algorithm to search all candidate surfaces (including quadric surface and plane surface). Furthermore, the proposed method can deal with the 3D wireframe models with no restriction on the axis of the quadric surface and identify the type of the quadric. 2011 Vol. 24 (3): 321-326 [Abstract] ( 822 ) [HTML 1KB] [ PDF 363KB] ( 620 ) 327 Factor Analysis Feature Extraction Algorithm Based on Shannon Entropy JIA Wei-Kuan, DING Shi-Fei, XU Xin-Zheng, SU Chun-Yang, SHI Zhong-Zhi- The performance assessments of existing data extraction algorithms mostly use variance contribution rate calculated by eigenvalues of raw data to measure the effect of feature extraction. However, variance contribution rate emphasizes the characteristic of eigenvalues of correlation matrix of the sample and it can not take information measuring into account. The extraction effect can be assessed from the angle of information theory by introducing Shannon information entropy into extraction algorithm, defining class probability and class information function and determining feature dimensions by calculating total information contribution rate. The theory are combined with factor analysis (FA) and FA feature extraction algorithm of information function is established. The extracting number of main factors is determined by information contribution rate. Finally, the efficiency of the theory is tested by cases. 2011 Vol. 24 (3): 327-331 [Abstract] ( 1079 ) [HTML 1KB] [ PDF 324KB] ( 690 ) 332 An Approach to Construction of Scenario Map in Chance Discovery Based on Six Degrees of Separation Theory ZHANG Zhen-Ya, CHENG Hong-Mei, ZHANG Shu-Guang The constructing of scenario map is one of the key activities in the process of chance discovery. The event correlation matrix is extended according to six degrees of separation theory and the formalization description of the structure of scenario map in chance discovery is presented. The principle for the construction of event clusters in scenario map via clustering analysis based on p order correlation matrix and the construction of scenario map based on event clusters are given. The algorithm for the implementation of the principle is given as well. The evaluating criterion for the performance of the proposed methods is discussed and the efficiency coefficient is used as the criterion definitely. Experimental results show that the clustering precision and the efficiency coefficient of scenario map constructed based on correlation matrix are high and the construction of scenario map based on 3order correlation matrix is fine for online chance discovery. 2011 Vol. 24 (3): 332-339 [Abstract] ( 688 ) [HTML 1KB] [ PDF 476KB] ( 557 ) 340 A Method for Multi Attribute Large Group Decision Making With Interval Number Information WAN Shu-Ping Aiming at multiattribute and multi alternative large group decision making in which the attribute values are in the form of interval numbers, an interval clustering algorithm is proposed. The attribute  values of alternatives are clustered to obtain the large group preference matrix of alternatives by the proposed algorithm. The weight vector of attributes is derived through integration of induced ordered weighted averaging (IOWA) operator. The result of alternative priority is given according to the overall evaluation values of all alternatives, thereby, the method of large group decision making is proposed. This method avoids the empirical selection of threshold value to cluster well and improves the credibility of decision making. The example analysis verifies that the method is both effective and practicable. 2011 Vol. 24 (3): 340-345 [Abstract] ( 772 ) [HTML 1KB] [ PDF 336KB] ( 615 ) 346 Dorsal Hand Vein Recognition Algorithm Based on Ridgelet Transforming of Divided Blocks JIA Xu, XUE Ding-Yu, CUI Jian-Jiang, LIU Jing A vein recognition algorithm based on fusing all local directional features which are extracted from divided blocks is proposed. Firstly, the acquired binary image is thinned by improved thinning algorithm after vein image preprocessing and the vein skeleton information is obtained. Secondly, the thinned vein image is divided into blocks. Then, every subimage is processed by ridgelet transforming, the dimensions of ridgelet transforming coefficients are reduced by applying principal component analysis, and the eigenvectors of vein image are acquired. Finally, vein images are classified and matched through making use of support vector machine based on the eigenvectors of image. Experimental results show that eigenvectors which are acquired through proposed algorithm have better discrimination, recognition results are affected less by errors that are generated in image acquiring and preprocessing, and the correct recognition rate exceeds 97%. 2011 Vol. 24 (3): 346-352 [Abstract] ( 785 ) [HTML 1KB] [ PDF 433KB] ( 603 ) 353 An Improved Gait Recognition Method Based on Ground Reaction Force LIN 尔Dong, YAO Zhi-Ming, ZHENG [Zhong, ZHOU Xu, SUN Xiang-Yang, SUN Yi-Ning An improved gait recognition approach based on ground reaction force (GRF) is proposed. 3directional GRF are acquired by 3dimensional force plate while a person is walking through the gait walkway. Wavelet packet (WP) decomposition is used to extract features in timefrequency domain, and optimal feature subset is selected using a fuzzy cmeans (FCM) clustering algorithm. Support vector machine (SVM) classifier is trained on trainingset, and then gait recognition is implemented by SVM on testing set. In order to improve the recognition accuracy, waveform alignment and resampling approach are utilized. Multiple classifiers are designed to reduce the negative influence of changes in walking speed. The approach is tested on a gait database collected from 103 subjects. Comparative results demonstrate that high recognition accuracy can be reached even in fewer trainingsamples. 2011 Vol. 24 (3): 353-359 [Abstract] ( 891 ) [HTML 1KB] [ PDF 424KB] ( 636 ) 360 Directional Filter Masks for Fingerprint Enhancement Based on Fibonacci Sequences CAI Xiu-Mei, FAN Jiu-Lun, GAO Xin-Bo The coefficients of the filter masks are often determined experientially or experimentally. Some of the filter coefficients are beyond the range of the mask after rotation, which destroys the distributed regularity of the coefficients. Aiming at this problem, the size of the filter mask is expanded to resolve the problem of overflow with coefficients distributed regularity maintained. On the other hand, a section of the famous 〖JP2〗Fibonacci sequence is chosen as the coefficients to reduce the influence of human factors.〖JP〗 The experimental results show that this directional filter mask works more effectively than the existing procedures. Moreover, it connects the breaks and separates conglutinated ridges simultaneously. 2011 Vol. 24 (3): 360-367 [Abstract] ( 733 ) [HTML 1KB] [ PDF 519KB] ( 720 ) 368 Efficient Image Categorization Based on Improved Distributions of Local Features GUO Li-Jun, LIU Xi, ZHAO Jie-Yu, SHI Zhong-Zhi An efficient image categorization method based on improved distribution of local invariant features is proposed. Firstly, the distribution of local features of an image is modeled by Gaussian mixed models (GMMs). Then, initial probability signatures are established by projecting local features onto all single models of GMMs. Finally, the complete probability signatures are obtained by performing a compression process. The probability signatures retain high discriminative power of probability density function (PDF) model, and they are suited for measuring dissimilarity of images with earth mover's distance (EMD), which allows for partial matches of compared distributions. The images are classified by learning support vector machine classifier with EMD kernel. The proposed method is evaluated on three image databases in scene recognition and image categorization tasks. The results of comparative experiments show that the proposed algorithm has inspiring performance. 2011 Vol. 24 (3): 368-375 [Abstract] ( 1163 ) [HTML 1KB] [ PDF 583KB] ( 850 ) 376 A Weighted Learning Vector Quantization Algorithm AN Xing, LIU Zhi-Wen, SHI Yong-Gang, 吕Chuan-Feng In view of the fact that difference of the importance of different dimensions is not taken into account in traditional learning vector quantization (LVQ) algorithms, a weighted LVQ algorithm is presented. In the proposedalgorithm, a set of additional weights is introduced for each neuron to indicate the importance of their respective dimensions.The weights are updated adaptively regarding the fitness of their corresponding neuron over the training iteration. The updating thresholds and step are decided according to the mean value of distance of all dimensions. Furthermore, according to the mean value of distance, it gets better stability and updates the weights without normalization. Six well known databases from UCI machine learning repository are selected to verify the performance of the proposed weighted LVQ (WLVQ) algorithm. The experimental results show that the proposed method gains insight to the role of the data dimensions, especially local weights, and yields the superior performance in recognition rate, stability, and computational complexity. 2011 Vol. 24 (3): 376-384 [Abstract] ( 812 ) [HTML 1KB] [ PDF 627KB] ( 639 ) 385 Web Information Extraction Based on Genetic Algorithm GUO Yin-Rui, CHEN Rong WHISK system is a semi automatic information extraction (IE) system. It works well in extracting information for structured or semi structured web texts. However, but there is no guarantee that the rule learning algorithm can extend rules in an optimal way. Besides, the generation of rule set is  time consuming. To solve these problems, the genetic algorithm is introduced to improve the supervised machine learning algorithm WHISK by a heuristic rule expansion, and a removing method is used to generate the rule set. The experimental results show that the proposed algorithm performs well in terms of the efficiency and the recall rate. 2011 Vol. 24 (3): 385-390 [Abstract] ( 665 ) [HTML 1KB] [ PDF 395KB] ( 594 ) 391 Quantum Immune Clonal Algorithm Based on Biologic Information Mechanism LIU Shi-Rong, ZHANG Bo-Tao The efficiency of immune clonal algorithm is greatly reduced by evolution randomness due to random mutations in polyclonal strategies. A quantum immune clonal algorithm based on biologic information mechanism (QICABIM) is proposed to solve the problem. In QICABIM, quantum computation is introduced into mutations of polyclonal strategies to facilitate genetic operation, and a biologic information mechanism is employed to improve information interaction and accelerate the speed of evolution. The convergence of QICABIM is then proved. Simulation results indicate that QICA BIM significantly improves the optimizing capability of immune clonal algorithm. Besides, it has better search capability and higher stability compared with some other quantuminspired immune clonal algorithms,advanced immune clonal algorithms and evolution algorithms. 2011 Vol. 24 (3): 391-399 [Abstract] ( 752 ) [HTML 1KB] [ PDF 544KB] ( 608 ) 400 Application of Case Based Reasoning in Cognitive Engine FENG Wen-Jiang, LIU Zhen, QIN Chun-Ling Genetic algorithm can be used to solve the transmission parameters optimization problem in cognitive engine. As the cognitive user increases, the ascending chromosome results in long convergence time of genetic algorithm, which can not satisfy the real time communication application of cognitive radio. According to the characteristic of existing cognitive engine, a cognitive engine architecture is proposed, and case based reasoning algorithm is introduced based on that. Then, casebased reasoning is utilized to search the matched cases. and provides genetic algorithm with initial population, which avoids the blindness of initial population searching. Simulation results show that compared with the general genetic algorithm, the cognitive engine with the proposed algorithm increases the convergence speed and utility function values and improves the cognitive engine capability significantly. 2011 Vol. 24 (3): 400-404 [Abstract] ( 654 ) [HTML 1KB] [ PDF 320KB] ( 520 ) 405 Nonlinear Radon Transform and Its Application to Face Recognition GAN Jun-Ying, HE Si-Bin Three nonlinear Radon transforms, including parabola, hyperbola, and ellipse Radon transform, are studied respectively, and the relationships among them are analyzed. Then, the three nonlinear Radon transforms are applied to face recognition. When shape parameters of parabola, hyperbola, and ellipse are approximately infinite, linear Radon transform of image is equal to parabola Radon transform and hyperbola Radon transform is equal to ellipse Radon transform. Nonlinear Radon transform possesses the characters of reducing noise and can be used to represent texture features of image. Moreover, polluted face images are represented by feature matrix via three nonlinear Radon transforms, and then combined with principal component analysis in face recognition. Experimental results demonstrate the validity of nonlinear Radon transform in face recognition. 2011 Vol. 24 (3): 405-410 [Abstract] ( 882 ) [HTML 1KB] [ PDF 344KB] ( 603 ) 411 A Hierarchical Algorithm for Human Posture Recognition Based on Spatial and Frequency Domain Features DENG Tian-Tian, WANG Zhi-Ling, ZHU Ming-Qing, CHEN Zong-Hai Human posture recognition is a research hotspot in the automatic video understating technology. It is difficulty to ensure the accuracy, robustness and realtime at the same time in practical applications. The existing mainstream algorithms based on 2D image information can be classified into two classes: the methods based on high level human model which have high accuracy and highcomplexity; the methods based on low level image information which have low complexity and low accuracy. A algorithm for human posture recognition is proposed to solve this problem. Firstly, Gaussian mixture model is exploited to extract foreground and normalized human silhouette. Then, a 12 dimensional invariant eigenvector is constructed, thereby the human posture model is established. Finally, a hierarchical recognition method is adopted to recognize the postures. This algorithm is efficient, has low complexity, and achieves good effect for some interfered images. The results of the experiments based on standard video database verify the validity of the proposed algorithm, and the superiority of the algorithm is also verified compared with chain code algorithm. 2011 Vol. 24 (3): 411-416 [Abstract] ( 648 ) [HTML 1KB] [ PDF 421KB] ( 948 ) 417 Bi Direction Weighted (2D)2 PCA with Eigenvalue Normalization One for Finger Vein Recognition GUAN Feng-Xu, WANG Ke-Jun, LIU Jing-Yu, MA Hui To carry out the finger vein recognition quickly and effectively, an algorithm of finger vein recognition is proposed according to the characteristics of bidirection two dimensional principal component analysis ((2D)2PCA) reducing the dimensions. The algorithm is bidirection weighted (2D)2PCA with eigenvalue normalization one ((OW2D)2PCA) based on preprocessing image of the figure vein image. The effect of the rate of cumulate eigenvalue on (2D)2PCA is analyzed, and the effect of the weighted value, the weighted value with eigenvalue normalization one and the rate of cumulate eigenvalue on W(2D)2PCA、OW(2D)2PCA、(W2D)2PCA and (OW2D)2PCA are analyzed as well. Experimental results on our database of finger vein images show that the presented method achieves high recognition accuracy. The redundant information of eigenvectors extracted by (2D)2PCA is restrained strongly, and the bi direction weighted effect is better than the one direction weighted effect. The average recognition rate of (OW2D)2PCA is higher than those of 2DPCA、(2D)2PCA、W(2D)2PCA、(W2D)2PCA and OW(2D)2PCA. 2011 Vol. 24 (3): 417-424 [Abstract] ( 725 ) [HTML 1KB] [ PDF 506KB] ( 655 ) 425 DTW Based Pattern Matching Method for Multivariate Time Series LI Zheng-Xin, ZHANG Feng-Ming, LI Ke-Wu Existing methods for matching multivariate time series can not measure similarity efficiently and accurately at the same time. Multivariate time series are fitted with multidimensional piecewise method. The angle of inclination and time span of a fitting line segment are chosen as feature pattern, and then a pattern matching method based on DTW for multivariate time series is proposed. Finally, its validity is testified by experiments. The experimental results show that the similarity of multivariate time series are measured efficiently and accurately by the proposed method, especially for series which present a whole process in a comparatively long time. 2011 Vol. 24 (3): 425-430 [Abstract] ( 966 ) [HTML 1KB] [ PDF 398KB] ( 844 ) 431 Area Coverage Algorithm in Swarm Robotics Based on Wasp Swarm Algorithm ZHANG Guo-You, ZENG Jian-Chao The definition, classification and application of coverage are summarized. Formalized definition of terrain coverage problem is discussed adopting triples notation. After briefly introducing the response threshold model of wasp swarm, the area coverage algorithm of swarm robots based on response threshold model during the terrain coverage process is depicted in detail. The analysis of the relationship of average moving probability, response threshold and stimulus are described to guide the selection of response threshold. Finally, the relationships of area coverage rate, duplicated coverage number of cell, response threshold and stimulus are discussed by simulation experiments, which are executed on different terrain scale, the number of robots and response thresholds. Simulation results show that the proposed algorithm is effective and feasible. 2011 Vol. 24 (3): 431-437 [Abstract] ( 738 ) [HTML 1KB] [ PDF 471KB] ( 1113 ) 438 Clonal Selection Algorithm Based on Grade Variation SONG Dan, LAI Xu-Zhi, WU Min A clonal selection algorithm is proposed, called clonal selection algorithm based on grade variation (CSABGV). To improve the effectiveness of the variation, the variability scale is divided into several levels: low grade variant is conducive to jump out of local optimal solution and to achieve global optimization while high grade variation is in favor of local optimization. In addition, the algorithm remembers and uses the information of variation grade of the parent antibody, and develops effective mutation strategy to guide mutation of antibodies. The performance of the proposed algorithm is compared with five benchmark functions and other optimization algorithms. Experimental results show that CSABGV has the characteristics of rapid convergence, powerful global search capability, high precision and good robustness. 2011 Vol. 24 (3): 438-443 [Abstract] ( 717 ) [HTML 1KB] [ PDF 417KB] ( 610 ) 444 One Class Classification Algorithm Based on Sparse Minimum Spanning Tree Adaptive Covering Model in HighDimensional Space HU Zheng-Ping, LU Liang, XU Cheng-Qian Minimum spanning tree class descriptor (MSTCD) describes the target class with the assumption that all the edges of the graph are basic elements of the classifier, which offers additional virtual training data for a description of sample distribution in highdimensional space. However, this descriptive model has too many branches, which results in the model being more complicated. According to the continuity law of the feature space of similar samples, a one class classification algorithm based on sparse minimum spanning tree covering model is presented. The method firstly constructs sparse  k nearestneighbor  graph representation for the target class. Then, a recursive graph bipartition algorithm is introduced to find the microcluster. Finally, it builds sparse minimum spanning tree on the graph nodes which are centers of micro cluster. Experimental results show that the presented algorithm performs better than MSTCD and other one class classifiers. 2011 Vol. 24 (3): 444-451 [Abstract] ( 677 ) [HTML 1KB] [ PDF 475KB] ( 788 ) 452 A Hybrid Constrained Semi Supervised Clustering Algorithm LI Xue-Mei, WANG Li-Hong, SONG Yi-Bin A hybrid constrained semi supervised clustering algorithm(HCC) is proposed based on consistency algorithm. To get a better clustering result, both labeled data and pairwise constraints are considered in clustering to make use of two types of prior knowledge supplementary to each other. The theoretical derivation and the algorithm are presented in detail. Experimental results show that labeled data outperform pairwise constraints in promoting the quality of clustering. Additionally, for many indices, such as CRI, number of clusters and running time, HCC is better than comparative algorithms. 2011 Vol. 24 (3): 452-456 [Abstract] ( 681 ) [HTML 1KB] [ PDF 274KB] ( 573 )

Supervised by China Association for Science and Technology Sponsored by Chinese Association of Automation NationalResearchCenter for Intelligent Computing System Institute of Intelligent Machines, Chinese Academy of Sciences Published by Science Press

Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn Supported by Beijing Magtech 　Email:support@magtech.com.cn