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.5, Published 2011-10-25 Orignal Article        Orignal Article 597 3D Model Retrieval with Multi-Granular Semantics Based on Gaussian Process Classifier GAO Bo-Yong, ZHANG San-Yuan, PAN Xiang In order to solve the inconsistency between users’ intentions in semantic 3D model retrieval system, a retrieval framework with multi-granular semantics is established, in which learning model can adapt to different user search intentions. Firstly, model classification is divided into different levels and the multi-granularity structure of semantic concept is formed. Then, a hybrid shape feature based on views is used to describe the shape characteristics of 3D model. And the Gaussian process classifier is used to associate low-level features with query concepts on a different level of semantic concept. Compared with existing research, the retrieval framework with multi-granular semantics allows the users to set their retrieval intentions according to selecting the granular level of semantics, and the results meet the user semantics as much as possible. The experimental results of retrieval performance evaluation using the benchmark show that the retrieval performance using proposed method is significantly higher than content-based retrieval and confident with human concept. 2011 Vol. 24 (5): 597-603 [Abstract] ( 607 ) [HTML 1KB] [ PDF 554KB] ( 600 ) 604 Automatic Image Segmentation Method Based on Graph Cuts GUO Bao-Long, HOU Ye Aiming at graph cuts majoring in interactive image segmentation, an automatic image segmentation method based on graph cuts is proposed. It can be used in both segmenting color images and segmenting gray images. In this method, the data item and the smooth item of the energy function are established after initialization. The energy function is solved by graph cuts. The series of steps are implemented iteratively until certain condition is met. The method does not need to make any constraints, build image model or estimate data distribution. It obtains good segmentation result rapidly. Gray images and color images are segmented by experiments. The experimental results show that the proposed approach is favorable. 2011 Vol. 24 (5): 604-609 [Abstract] ( 717 ) [HTML 1KB] [ PDF 419KB] ( 806 ) 610 Automata and Grammars Theory Based on Complete Residuated Lattice-Valued Logic Peng Jia-Yin A fundamental framework of automata and grammars theory based on complete residuated lattice-valued logic is preliminarily established. Firstly, the concept of l value regular grammars is introduced. It is proved that any l value language recognized by l value automaton is equivalent to that generated by some l value regular grammar, and conversely, the l value language generated by any l value regular grammar is also equivalent to that recognized by some l value automaton. Afterwards, the concatenations of l value automaton and l value language recognized by l value automaton are depicted. In particular, the l value pumping lemma and L value pumping lemma are built, and then a decision characterization of l value language is presented. Finally, the equivalence between the l value automata with ε-transitions and those without ε-transitions is revealed. 2011 Vol. 24 (5): 610-618 [Abstract] ( 500 ) [HTML 1KB] [ PDF 437KB] ( 742 ) 619 Keyframe Extraction from Human Motion Capture Databy Simplex Hybrid Genetic Algorithm LIU Xian-Mei, HAO Ai-Min, ZHAO Dan To obtain a compact representation of human motion based on keyframes, a method for keyframes extracting of the captured human motion data by simplex hybrid genetic algorithm is presented, which combines genetic algorithm with a local search technique to converge faster and produce the optimal solution. Firstly, the fitness function is defined to evaluate the availability of keyframe with the goals of minimal reconstruction error and optimal compression rate. Then, the reconstruction error is computed between the original motion and the reconstruction one by the weighted differences of joint positions and velocities. The velocity term helps to preserve the dynamics of motion. Finally, the individuals of initial population are optimized by the knowledge to assure the evolutionary efficiency and the population diversity. Experimental results show that the proposed method can effectively extract keyframes, produce remarkable results in terms of quality and compression ratio, and reconstruct all other non-keyframes of an animation with these keyframes. 2011 Vol. 24 (5): 619-628 [Abstract] ( 591 ) [HTML 1KB] [ PDF 732KB] ( 683 ) 629 Manifold Outlier Detection Algorithm Based on Local-Correlation Dimension HUANG Tian-Qiang, LI Kai, GUO Gong-De Traditional outlier detection algorithm is not suitable for detection of manifold outlier. There are reports of denoising algorithm for manifold learning, but fewer reports of manifold outlier detection algorithms. Therefore, the manifold outlier detection algorithm is proposed based on the local-correlation dimension according to experimental observations. Firstly, the nature of the intrinsic dimension is discussed, and the local-correlation dimension is used to measure the manifold outlier, which is based on experimental observations. And then it is proved that the nature of outliers on manifolds can be characterized by local-correlation dimension. Finally, the manifold outlier detection algorithm based on local-correlation dimension is proposed according to the nature. The performance evaluation of the artificial data and the real data shows that the algorithm can detect manifold outliers and it has better performance than the recently reported manifold blurry mean shif algorithm. 2011 Vol. 24 (5): 629-636 [Abstract] ( 622 ) [HTML 1KB] [ PDF 482KB] ( 681 ) 637 Clustering Method Based on Structural Similarityand Compressive Transformation MOU Lian-Ming, ZHAN De-Chuan, LI Ming, ZHOU Zhi-Hua The current clustering methods are difficult to handle the complicated problems in which shapes and densities are changing along with the data. To overcome the shortcomings of existing clustering methods, based on discrete topological manifold created in the data space, the structural similarity of samples and the class structure are described by accessibility after defining two new relativity metrics: the neighborhood density similarity and the smoothness of neighborhood density changes. The class structure relationship is proved to an equivalence relation. Then, a clustering algorithm is designed based on compressive transformation by treating the structural similarity defined on samples as the attractiveness. The algorithm is designed to handle data with any shapes and any density, maintaining good interpretability and many other advantages. Experimental result on the artificial data sets and standard data sets shows that the method is superior to the state-of-the-art methods. 2011 Vol. 24 (5): 637-644 [Abstract] ( 288 ) [HTML 1KB] [ PDF 532KB] ( 619 ) 645 A Learning Algorithm for Shortest Branch Cut Length Problem ZHENG Dong-Liang, DA Fei-Peng Branch cut method is an effcient noise-immune algorithm for correct phase unwrapping of noisy phase maps. The shortest branch cut length promises the optimal unwrapping of the wrapped phase maps. The shortest branch cut length problem belongs to combinatorial optimizations. A learning algorithm is proposed to resolve the problem. One solution for the problem is one individual for the algorithm. Individuals learn from other individuals and mutate by themselves to realize the evolution, which is similar to the crossover and mutation operator in the genetic algorithm. Compared with the traditional methods, the learning algorithm is fast and competitive. 2011 Vol. 24 (5): 645-650 [Abstract] ( 481 ) [HTML 1KB] [ PDF 442KB] ( 646 ) 651 Track Initiation Algorithm Based on Randomized Hough Transform SHI Yin-Shui, JI Hong-Bing, WANG Xue-Qing, Cui Xun-Xue Radar usually needs to detect targets under the dense pulse jamming condition, and the target information mixes up the fragmentary plots of false target jamming after it has been jammed by those pulses. It is difficult for the existing methods to initiate tracks effectively under the false target jamming condition. A track initiation algorithm based on the traditional randomized Hough transformation is proposed. The technology of sequence check is adopted to design the rule of sampling termination. Moreover, the parameter set of a flight feature constraint projection is constructed to eliminate any false tracks. Simulation results show that the proposed algorithm provides an excellent practicability for an information synthesis system, its average ratio of false tracks is only 14.9%,and its computation time occupies 9.5416 seconds. 2011 Vol. 24 (5): 651-657 [Abstract] ( 360 ) [HTML 1KB] [ PDF 440KB] ( 616 ) 658 Graph-Optimized Linear Discriminant Projection and Its Application to Image Recognition YIN Jun, JIN Zhong The class information of the data is sufficiently utilized and a feature extraction algorithm is proposed called graph-optimized linear discriminant projection (GoLDP) based on graph-optimized locality preserving projection (GoLPP). The graph of GoLDP is constructed by optimizing an objective function, which is similar to GoLPP. GoLDP constructs two optimal graphs (optimal intrinsic graph and optimal penalty graph) by using class information, which is different from GoLPP, and obtains the optimal projection matrix according to these two optimal graphs. Experimental results on FERET and YALE face databases and the PolyU palmprint database demonstrate the effectiveness of GoLDP. 2011 Vol. 24 (5): 658-664 [Abstract] ( 437 ) [HTML 1KB] [ PDF 443KB] ( 631 ) 665 Symbolic Aggregate Approximation Based on Shape Features LI Hai-Lin, GUO Chong-Hui Changeable trends of time series can be reflected by shape features which retain sufficient data information during the dimensionality reduction. It is good to improve the efficiency of time series data mining in the later stage. A symbolic aggregate approximation based on shape features is proposed. It regards the mean and the shape feature of a sequence as two important characteristics, and changes their domains of discourse to transform them into strings. Compared with the traditional methods, the proposed method improves the efficiency of time series data mining in the setting of equal compress rate because of the sufficient information which is retained by the previous stage. 2011 Vol. 24 (5): 665-672 [Abstract] ( 395 ) [HTML 1KB] [ PDF 529KB] ( 1047 ) 673 A Survey of Context Reasoning Methods in Ambient Intelligence LIU Da-You, LIU Chun-Chen, WANG Sheng-Sheng Context reasoning is a key topic in Ambient Intelligence (AmI), which impacts on the intelligent, sensitive, responsive and adaptive ability of AmI systems, and it has gained much attention from researchers in the recent years. The primary research contents, methods and advancements of context reasoning are proposed and analyzed. Exist problems and research directions are also discussed. 2011 Vol. 24 (5): 673-679 [Abstract] ( 471 ) [HTML 1KB] [ PDF 585KB] ( 754 ) 680 An Artificial Glowworm Swarm Optimization Algorithm Based on Powell Local Optimization Method ZHANG Jun Li, ZHOU Yong Quan In order to overcome the shortcomings of artificial glowworm swarm optimization (GSO) algorithm including slow convergence speed, easily falling into local optimum value, low computational accuracy and low success rate of convergence, an artificial GSO algorithm based on Powell local optimization method is proposed. It adopts the powerful local optimization ability of Powell method and embeds it into GSO as a local search operator. Experimental results of 8 typical functions show that the proposed algorithm is superior to GSO in convergence efficiency,computational precision and stability. 2011 Vol. 24 (5): 680-684 [Abstract] ( 315 ) [HTML 1KB] [ PDF 291KB] ( 667 ) 685 Optimizing Data-Dependent Kernel Using Semi-supervised Learning with Pairwise Constraints WANG Na, LIU Guo-Sheng, LI Xia The selection of kernel function and its parameters determine the performance of kernel function. A semi-supervised data-dependent kernel optimization algorithm is presented, which uses unlabeled data and pairwise constraints to maximize an objective function sensitive to data-dependent kernel, so that its performance is improved. Then the proposed method is employed to optimize the kernel of kernel principal components analysis (KPCA) and the experimental results of the classification and clustering performance on the artificial data and UCI data sets show its efficiency. 2011 Vol. 24 (5): 685-691 [Abstract] ( 452 ) [HTML 1KB] [ PDF 425KB] ( 562 ) 692 A Glasses-Occluded Face Recognition Method Based on 3D Face Reconstruction XIONG Peng-Fei, LIU Chang-Ping, HUANG Lei The instability of eyeglasses treated as facial features is the primary obstacle in glasses-occluded face recognition. The facial features are easily lost by the existing method while eliminating the influence of instable glasses characteristics. To avoid this problem, a 3D realistic face model reconstruction is applied for virtual face images generation to compensate the instability of the glasses. In the method, the glasses are set as an inherent part of face. 3D face reconstruction increases the feasibility of parameter adjustment for different glasses models. Based on this, various influences of glasses segment on face recognition are analyzed. Also, corresponding solutions to image distortion by lens blur and reflection are carried out. Experiments on CAS-PEAL database demonstrate the improvement of the proposed method on face recognition rate and verify the effectiveness of lens treatment. 2011 Vol. 24 (5): 692-699 [Abstract] ( 342 ) [HTML 1KB] [ PDF 613KB] ( 933 ) 700 Identification of Lung Nodules in CT Images Based on 3D Minimum Within-Class Scatter SVM WANG Qing-Zhu, KANG Wen-Wei, WANG Xin-Zhu, Wang Bin Multi-class Support Vector Machines (MC-SVM) based on 3 Dimension (3D) minimum within-class scatter is presented. MC-SVM based on 3D matrix patterns (MC-SVM3Dmatrix) is proposed firstly which operates inputs as 3D matrixes directly, and minimum within-class scatter SVM is adopted to design a 3D minimum within-class scatter MC-SVM. Taking advantages of both minimum within-class scatter SVM and feature of 3D space, the algorithm improves accuracy of classifiers and reduces False Positives (FP) effectively. 200-case database from Jilin Tumor Hospital is used to validate the proposed algorithm. The performances of other four CAD schemes, two radiologists and the proposed algorithm are compared on the same database. The experimental results verify the effectiveness of the proposed algorithm. 2011 Vol. 24 (5): 700-706 [Abstract] ( 314 ) [HTML 1KB] [ PDF 447KB] ( 713 ) 713 Grid Computing Workflow Scheduling Clonal Selection Algorithm with Multi-QoS Constraints ZHAO Jian-Feng, ZENG Wen-Hua, LIU Min, ZHANG Xue Workflow scheduling with multi-QoS constraints is hard to be solved under the grid computing environment. A clonal selection algorithm, named EvoWF, is proposed to solve workflow scheduling problem based on deep analysis on the difficulty of this problem. The encoding of working scheduling is simplified by adding the grid service identification. The concept of QoS preference is proposed, which converts object function of workflow scheduling to fitness function, and QoS attributes can be extended. Compared to genetic algorithm and ant colony optimization, EvoWF is more efficient. In extension, EvoWF gets the same optimum scheduling results compared with the single-QoS constraint greed time or cost algorithm. Moreover, the effect of parameters is analyzed by experiments. 2011 Vol. 24 (5): 713-724 [Abstract] ( 331 ) [HTML 1KB] [ PDF 570KB] ( 811 )

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