ISSN 0254-4156 CN 11-2109/TP   　·About Journal 　·Editorial Board 　·News Center 　　Conference 　　R&D Reports 　　Other News 　　Related Journals 　·Journal 　　Current Issue 　　Archive 　　Forthcoming Articles 　　Advanced Search 　　Browse by 　·Editorial Office 　　About us 　　Editing Work 　　Provide Information 　　Bug Report 　·Editorial Notice · Call for papers: Sep...(2008-10-30)(New) 　　　　　　　　More...   Total visitors:   Now online: sunglasses     Latest Printed Issue   2012  Vol.38 Number 5 May 20, 2012 Content is this issue   综述   Papers and Reports   Brief Papers         综述     673 XU Xin, SHEN Dong, , GAO Yan-Qing, , WANG Kai and   Learning Control of Dynamical Systems Based on Markov Decision Processes: Research Frontiers and Outlooks     Learning control of dynamical systems based on Markov decision processes (MDPs) is an interdisciplinary research area of machine learning, control theory, and operations research. The main objective in this research area is to realize data-driven multi-stage optimal control for complex or uncertain dynamical systems. This paper presents a comprehensive survey on the theory, algorithms, and applications of MDP-based learning control of dynamical systems. Emphases are put on recent advances in the theory and methods of reinforcement learning (RL) and adaptive/approximate dynamic programming (ADP), including temporal-difference learning theory, value function approximation for continuous state and action spaces, direct policy search, approximate policy iteration, and adaptive critic designs. Applications and the trends for future research and developments in related fields are also discussed.     2012 Vol. 38 (5): 673-687 [Abstract ] ( 76 ) [HTML 0KB] [PDF 767KB ]( 89 ) 688 ZHANG Su-Lan, , GUO Ping, ZHANG Ji-Fu and HU Li-Hua   Automatic Semantic Image Annotation with Granular Analysis Method     To bridge the semantic gap between low-level visual feature and high-level semantic concepts has been the subject of intensive investigation for years in order to improve the accuracy of automatic image annotation and satisfy the users' needs of quick image retrieval. Granular analysis is a hierarchical and important data analyzing method, which provides a new idea and method for solving the complicated problem. The accuracy of automatic image annotation and the efficiency of image retrieval are varying with the granularity size of image understanding and analysis. In this paper, the state-of-art models of automatic semantic image annotation are overviewed, then the idea and models of the granular analysis with its application in the process of automatic semantic image annotation are discussed, and the granular analysis based automatic image annotation methods are investigated as well as the promising research directions are given.     2012 Vol. 38 (5): 688-697 [Abstract ] ( 42 ) [HTML 0KB] [PDF 584KB ]( 57 ) Papers and Reports     698 SUN Ming-Xuan and BI Hong-Bo   Learning Identification: Least Squares Algorithms and Their Repetitive Consistency     This paper presents a learning identification method for stochastic systems with time-varying parametric uncertainties. The systems undertaken perform tasks repetitively over a pre-specified finite-time interval, and a least squares learning algorithm is derived on the basis of the repetitive operations. The learning identification method applies to periodically time-varying systems. It is shown that the estimates converge to the time-varying values of the parameters, and the complete estimation can be achieved under repetitive persistent excitation condition, a sufficient condition for establishing repetitive consistency of the learning algorithms. Numerical results are presented to demonstrate the effectiveness of the proposed learning algorithms.     2012 Vol. 38 (5): 698-706 [Abstract ] ( 51 ) [HTML 0KB] [PDF 1050KB ]( 69 ) 707 MO Hong and WANG Tao   Computing with Words in Generalized Interval Type-2 Fuzzy Sets     The conventional fuzzy sets are type-1 fuzzy sets whose point-values are two-dimensional, and the point-values of type-2 fuzzy sets (T2 FSs) are three-dimensional. So it is more difficult for T2 FS to be understood and computed than the corresponding type-1. To make T2 FS be better understood and extensively applied, in this paper, we present the definition of generalized interval type-2 fuzzy sets (GIT2 FSs), and divide them into three types: discrete type, partial discrete type, and continuous type. Then, the corresponding mathematical representation of every type is given to get the extension principle formula of GIT2 FS. Lastly, computing methods are proposed to discuss computing with words of GIT2 FS by two different fuzzy logic operators.     2012 Vol. 38 (5): 707-715 [Abstract ] ( 44 ) [HTML 0KB] [PDF 1421KB ]( 51 ) 716 YANG Ren-Ming, and WANG Yu-Zhen   Stability Analysis and Estimate of Domain of Attraction for a Class of Nonlinear Time-varying Delay Systems     The stability and estimate of domain of attraction are studied for a class of nonlinear time-varying delay systems. Firstly, an equivalent form is obtained for the systems by means of coordinate transformation and orthogonal decomposition of vector fields. Then, based on the orthogonal condition and the free-weighting matrix method, several less conservative results are derived on the stability and estimate of domain of attraction. Finally, illustrative examples show effectiveness of the proposed method.     2012 Vol. 38 (5): 716-724 [Abstract ] ( 42 ) [HTML 0KB] [PDF 575KB ]( 56 ) 725 LIU Zhi-Yong   Graph Matching: a New Concave Relaxation Function and Algorithm     Recently, approximate graph matching based on relaxing the permutation matrix to a doubly stochastic matrix has become an important and popular topic. The key point lies in which approximation over a continuous set is usually easier to implement than that over a discrete one. However, a consequent trouble related to such a relaxation is how to properly map the doubly stochastic matrix back to a permutation one. In the literature, a concave relaxation function for matching problem between the undirected graphs without self-loops was recently proposed, such that the doubly stochastic matrix can converge to a permutation one in a smooth way, and got a state-of-art performance on matching accuracy. Unfortunately, except for the undirected graphs without self-loops, there are no concave relaxation proposed for any other types of graph models. In this paper, we propose a concave relaxation for the directed graphs without self-loops, based on which a graph matching algorithm is then presented. Extensive experimental comparisons witness the validity of the proposed methods.     2012 Vol. 38 (5): 725-731 [Abstract ] ( 41 ) [HTML 0KB] [PDF 517KB ]( 34 ) 732 LI Yi, SUN Zheng-Xing, CHEN Song-Le and LI Qian   3D Human Pose Analysis from Monocular Video by Simulated Annealed Particle Swarm Optimization     In this paper we proposed a simulated annealing particle swarm optimism (SAPSO) based method for human pose estimation form monocular image sequences. First, we use principle component analysis (PCA) to learn the low-dimensional compact space of human pose, by which the aim of both reducing dimensionality and extracting the prior knowledge of human motion are achieved simultaneously. Pose is estimated on the compact subspace. In the optimizing step, we introduce particle swarm optimism to human pose estimation, and further, a SAPSO pose estimation method is proposed. And last we use SAPSO to estimate and track human pose in monocular videos separately. Experimental results demonstrate that the proposed method is more convergent and globally optimum, which can estimate and track human pose in monocular images effectively.     2012 Vol. 38 (5): 732-741 [Abstract ] ( 37 ) [HTML 0KB] [PDF 2997KB ]( 35 ) 742 ZHU Hai-Long, LIU Peng, LIU Jia-Feng and TANG Xiang-Long   A Graph Analysis Method for Abnormal Crowd State Detection     An abnormity detection method for a dynamic crowd scene is proposed based on graph analysis. After the non-parametric clustering in velocity space via an adaptive mean shift algorithm, we get the clustering results containing some cluster centers and Euclidean distances between them, and they can form a graph whose vertexes are the cluster centers and edge weights are the distances. Through analyzing the vertexes' distribution in feature space and the state transform of a dynamic system made by the sequence of the edge weight matrix, we can detect and locate the abnormal events in the scenario. To testify the method's effectiveness, we conducted experiments on several well-known datasets and obtained good performance in both abnormal events detection and location. The results show that the graph analysis method has strong adaptability and can efficiently detect the abnormal states in dynamic crowd scene.     2012 Vol. 38 (5): 742-750 [Abstract ] ( 39 ) [HTML 0KB] [PDF 3883KB ]( 33 ) 751 LI Deng-Wang, LI Hong-Sheng, WANG Hui, WANG Hong-Jun, YIN Yong, and PENG Yu-Hua   Deformable Registration Method Using Edge Preserving Scale Space withApplication in Adaptive Radiation Therapy     Registration of planning CT with daily cone beam CT (CBCT) images is an important component for adaptive radiation therapy in CBCT based image guided radiation therapy (IGRT) system. The edge preserving scale space which is derived from the anisotropic diffusion model can provide abundant spatial information for mutual information based registration. For improving the registration efficiency in system, a multi-scale deformable registration framework is proposed by combining edge preserving scale space with the multi-level free form deformation (FFD) grids. Different scales use different FFD grid, where the deformation fields are gained by a coarse to fine manner. Furthermore, considering clinical application, we design an optimal method for estimation of the parameters in anisotropic diffusion model for automated registration. The experiment results demonstrate that the proposed method can register the daily CBCT with the planning CT accurately and has a promising efficiency when used in CBCT based IGRT system. After gaining the deformation field, tumor target, organ at risk (OR) and iso-dose are contoured automatically, and then ''DVH (Dose volume histograms) analysis'' are also studied. Finally, radiation planning is transferred from planning CT to daily CBCT images adaptively.     2012 Vol. 38 (5): 751-758 [Abstract ] ( 44 ) [HTML 0KB] [PDF 6163KB ]( 25 ) 759 ZHU Xiong-Yong, ZHOU Jie and TAN Hong-Zhou   Method for Eliminating LCD Motion De-blurring Model's Pole     The sinc-1 model can reduce the LCD motion blur effectively, however, it can not restore the correct frequency components on the pole and is hard to implement on hardware. This paper presents a novel method which is trying to improve the LCD motion blur phenomenon from the view of system identification. The method uses the Volterra nonlinear system that is based on the VSSLMS (variable step size least mean square) algorithm to fit the sinc-1 model, which is a non-perfect inverse system to the LCD motion blur system for the pole problem. Simulation results demonstrate that the performance of the proposed simple method can avoid the pole problem of sinc-1 model and can be implemented on hardware easily.     2012 Vol. 38 (5): 759-768 [Abstract ] ( 36 ) [HTML 0KB] [PDF 3271KB ]( 24 ) 769 HU Zheng-Ping and FENG Kai   An Adaptive One-class Classification Algorithm Based on Multi-resolution Minimum Spanning Tree Model in High-dimensional Space     The coverage radii in the algorithm of MSTCD (Minimum spanning tree class descriptor) are generally fixed without diversification, which makes it difficult to construct a close coverage for the local structure. This article combines multi-resolution thought and minimum spanning tree (MST) covering model, and proposes an adaptive multi-resolution covering model based MST in high-dimensional space. In this algorithm, multi-resolution of data is determined by the distributed feature of data manifold itself. The resolution of any location is depended on the structure of sample point and edge. The proposed novel algorithm permits that the whole covering model could have different coverage radii or resolutions and the resolution has relationship with location. Experiments show that the algorithm is reasonable.     2012 Vol. 38 (5): 769-775 [Abstract ] ( 38 ) [HTML 0KB] [PDF 516KB ]( 25 ) 776 HE Chun, YE Yong-Qiang, JIANG Bin and ZHOU Xin   A Novel Edge Detection Method Based on Fractional-order Calculus Mask     In this paper, a new edge detection operator based on a composite derivative is proposed, which is realized by the combination of fractional differentiation and integration. The new complex edge-detection mask is also deduced for the implementation of fractional differentiation. The abilities of the compound derivative in terms of approximate simulation of first-order derivative and suppression of noise are demonstrated through one-dimensional examples. The experimental results of two-dimensional examples indicate that without the contamination of noise, the new operator can accurately detect the edge, while with the noise contamination, the new operator can effectively suppress the noise. Finally, quantitative analysis of the new operator is given. The results of the comparison between the new operator and Canny operator show that the new operator has the advantage of a low mispositioning rate.     2012 Vol. 38 (5): 776-787 [Abstract ] ( 40 ) [HTML 0KB] [PDF 2926KB ]( 32 ) 788 HUANG Chen, , , DING Xiao-Qing, , , FANG Chi, and   A Robust and Efficient Facial Feature Tracking Algorithm     Facial feature tracking obtains precise information of facial components in addition to the coarse face position and moving track, and is important to computer vision. The active appearance model (AAM) is an efficient method to describe the facial features. However, it suffers from the sensitivity to initial parameters and may easily be stuck in local minima due to the gradient-descent optimization, which makes the AAM based tracker unstable in the presence of large pose, illumination and expression changes. In the framework of multi-view AAM, a real time pose estimation algorithm is proposed by combining random forest and linear discriminate analysis (LDA) to estimate and update the head pose during tracking. To improve the robustness to variations in illumination and expression, a modified online appearance model (OAM) is proposed to evaluate the goodness of AAM fitting, then the appearance model of AAM is updated adaptively using the incremental principle component analysis (PCA). The experimental results show that the proposed algorithm has both efficiency and robustness.     2012 Vol. 38 (5): 788-796 [Abstract ] ( 46 ) [HTML 0KB] [PDF 3920KB ]( 37 ) 797 TAO Jian-Wen, and WANG Shi-Tong   Kernel Support Vector Machine for Domain Adaptation     Domain adaptation learning is a novel effective technique to address pattern classification, in which the prior information for training a learning model is unavailable or insufficient. To minimize the distribution discrepancy between the source domain and target domain is one of the key factors. However, domain adaptation learning may not work well when only considering to minimize the distribution mean discrepancy between source domain and target domain. In the paper, we design a novel domain adaptation learning method based on structure risk minimization model, called DAKSVM (kernel support vector machine for domain adaptation) with respect to support vector machine (SVM) and least-square DAKSVM (LSDAKSVM) with respect to least-square SVM (LS-SVM), respectively to effectively minimize both the distribution mean discrepancy and the distribution scatter discrepancy between source domain and target domain in some reproduced kernel Hilbert space, which is then used to improve the classification performance. Experimental results on artificial and real world problems show the superior or comparable effectiveness of the proposed approach compared to related approaches.     2012 Vol. 38 (5): 797-811 [Abstract ] ( 44 ) [HTML 0KB] [PDF 6212KB ]( 44 ) 812 YANG Bo, , LIU Jie, , LIU Da-You and   A Random Network Ensemble Model Based Generalized Network Community Mining Algorithm     According to the attributes of nodes and the linkages between them, most real-world complex networks could be assortative and disassortative. Community structures are ubiquitous in both types of networks. The ability to discovery meaningful community structures from both types of networks is fundamental for theoretical research and practical applications. Since the types of exploratory networks to be processed are usually unknown beforehand, it is difficult to determine what specific algorithms should be applied to them to obtain meaningful community structures. To address this issue, a novel concept of generalized network community is proposed in order to unify two concepts of assortative and disassortative communities. Based on a random network ensemble model, a generalized community mining algorithm, called G-NCMA, is proposed. Experimental results demonstrate that the G-NCMA algorithm is able to properly mine potential communities from explorative networks, as well as to determine their respective types.     2012 Vol. 38 (5): 812-822 [Abstract ] ( 38 ) [HTML 0KB] [PDF 3619KB ]( 27 ) 823 YANG Yi, HAN De-Qiang and HAN Chong-Zhao   Evidence Combination Based on Multi-criteria Rank-level Fusion     Dempster-Shafer evidence theory has been widely used in many important applications in information fusion, but Dempster's rule of combination always brings some counter-intuitive behaviors, e.g., the paradoxes of conflict, belief transfer, and belief absorbtion. Accordingly, a novel evidence combination approach based on multi-criteria rank-level fusion is proposed in this paper. It uses the criteria of evidence precision, evidence credibility, and evidence auto-conflict together to evaluate all the bodies of evidence to be combined. The combination result can be obtained based on selective fusion. The experimental results and related analysis show that the proposed approach is rational and effective.     2012 Vol. 38 (5): 823-831 [Abstract ] ( 34 ) [HTML 0KB] [PDF 1284KB ]( 23 ) 832 WEN Cheng-Lin and HU Yu-Cheng   Fault Diagnosis Based on Information Incremental Matrix     Principal component analysis (PCA) is a kind of commonly used fault detection method, but because of the uncorrected feature extraction, there are higher rates of false and missed alarm by using it in fault diagnosis. Thus, this paper firstly introduces the method of fault diagnosis based on the information incremental matrix obtained by the global covariance matrix. It can effectively reduce the rate of false and missed alarm as compared to PCA. But when the number of samples increases, the calculated threshold value is more unrepresentative and a much larger amount of calculation is required, they influence the performance of this method. Then, in order to overcome these shortcoming of the above method, one new fault diagnosis method is proposed by the local information incremental matrix obtained by the covariance matrix of moving the window, which comprises partial samples. This new method is mainly composed of defining the local covariance matrix, calculating local information incremental matrix, local information incremental mean, local dynamic threshold, and detecting abnormity and diagnosing fault, and so on. Finally, through two examples of numerical simulation to verify the detection efficiency of three fault diagnosis methods, i.e., PCA method, the method of fault diagnosis based on the information incremental matrix obtained by the global covariance matrix, and the proposed method, in false and missed alarm. The results show that the new method possesses the best detection performance.     2012 Vol. 38 (5): 832-840 [Abstract ] ( 47 ) [HTML 0KB] [PDF 1007KB ]( 44 ) 841 CHEN Hua, , ZHANG Jing, , ZHANG Xiao-Gang and HU Yi-Han   A Robust-ELM Approach Based on Parzen Windiow's Estimation for Kiln Sintering Temperature Detection     To eliminate the interference in the blurring pulverized coal flame image sequences of rotary kiln, a new kiln sintering temperature measurement method based on statistical features of pulverized coal flames and robust extreme learning machine (robust-ELM) is proposed in this paper. The degree of stability and quantity of radiant energy are computed from a blurry flames image sequences as statistical features, robust-ELM is presented to estimate the sintering temperature based on the above features of flames image. The distribution of training error of extreme learning machine (ELM) is estimated by Parzen windows to make up the weighted matrix to reduce the disturbance of gross errors in industrial field. Finally, a series of tests were undertaken on an industrial-scale flame videos, which showed the methods could measure sintering temperature more accurately, quickly, and robustly.     2012 Vol. 38 (5): 841-849 [Abstract ] ( 39 ) [HTML 0KB] [PDF 1673KB ]( 31 ) 850 ZHANG Feng, LUO Li-Min and BAO Xu-Dong   Automatic Diagnosis and Complete Parameters Extraction Algorithm for Breast Carcinoma Based on Electrical Impedance Scanning     As a non-invasive functional imaging method, electrical impedance scanning (EIS) has become an adjunctive diagnostic tool for X-ray mammography in diagnosis of breast carcinoma. EIS diagnostic method used by clinicians often has a large variability of sensitivity and specificity among different research groups. In this paper, we present a new parameter extraction method — complete parameters extraction algorithm (CPEA) and design a diagnostic indicator — abnormal energy indicator (AEI) for breast carcinoma, which could integrate diagnosis and parameter extraction into one process. Clinical experiments indicate that the method can be applied to diagnosis of breast carcinoma with higher sensitivity and specificity in a real-time way, and presents complete parameters of breast cancerous lesions simultaneously.     2012 Vol. 38 (5): 850-857 [Abstract ] ( 33 ) [HTML 0KB] [PDF 2165KB ]( 20 ) 858 LI Juan, , ZHAO You-Gang, YU Yang, ZHANG Peng and GAO Hong-Wei   Optimal Fault Diagnosis for Networked Control Systems withLarge Time-delays and Noises     The problem of fault diagnosis is investigated for networked control systems (NCS) with large measurement delays and noises. Based on the delay-free transformation approach, a novel design approach to optimal fault diagnoser is proposed. This scheme first establishes an augmented system with hidden fault states, and the networked control system with delays is transformed into a delay-free one by using the delay-free transformation approach. Then, a diagnosability criterion of faults is given. Furthermore, the design problem of the optimal fault diagnoser is transformed into the design problem of a state feedback controller by utilizing the duality principle. Finally, the real-time fault diagnosis is realized by constructing a novel optimal fault diagnoser which meets the quadratic performance index. The feasibility and validity of the proposed scheme are demonstrated by a simulation example.     2012 Vol. 38 (5): 858-864 [Abstract ] ( 44 ) [HTML 0KB] [PDF 1085KB ]( 31 ) Brief Papers     865 ZHAO Cong-Ran and XIE Xue-Jun   Output-feedback Regulation of Nonlinear Systems with iISS Inverse Dynamics     This paper discusses output-feedback regulation for more general nonlinear systems with integral input-to-state stability (iISS) inverse dynamics and unknown control direction. By using the adaptive backstepping method, an output feedback controller is given to drive the output to the origin while maintaining other closed-loop signals bounded.     2012 Vol. 38 (5): 865-869 [Abstract ] ( 32 ) [HTML 0KB] [PDF 279KB ]( 43 ) 870 MA Ru-Ning, TU Xiao-Po, DING Jun-Di and YANG Jing-Yu   To Evaluate Salience Map towards Popping out Visual Objects     The aim of this paper is to present a quantitative evaluation of five popular salience maps for object segmentation. First, five salience maps are revisited in terms of theory foundation. Second, human segmentation is taken as the ground truth of interesting objects ''pop-out'' to build three quantitative evaluation ratios of salience map to human segmentation. Finally, evaluation experiments are conducted on three image databases of Corel, MSRA and Weizmann. Results show some insights into the performances of these different salience maps in object segmentation. This research is believed meaningful and useful for the further development of salience-driven methods for object segmentation.     2012 Vol. 38 (5): 870-876 [Abstract ] ( 40 ) [HTML 0KB] [PDF 1095KB ]( 40 ) 876 SONG Yang, , DONG Hao, FEI Min-Rui and   Mean Square Exponential Stabilization of Markov Networked Control Systems Based on Switching Frequentness     This paper considers the mean square exponential stabilization for networked control systems (NCSs) based on a class of Markov models. Firstly, networked control systems are modeled as a discrete-time switched system of which the switching sequence is governed by a Markov chain with known transition probability matrix, and then a method is proposed to calculate switching frequentness. By combining the stochastic process theory as well as the switched systems stability theory, a mean square exponential stabilization method is presented. The state feedback control law can be obtained by solving a set of linear matrix inequalities. Finally, a numerical example is provided to demonstrate the effectiveness of the proposed result.     2012 Vol. 38 (5): 876-881 [Abstract ] ( 36 ) [HTML 0KB] [PDF 474KB ]( 32 ) 882 CHEN Gang and YU Ming   Synchronizing Control and Analysis of Distributed Passive Systems     Based on Lyapunov methods, a systematic way is proposed to design and analyze the synchronizing control protocols for the distributed passive systems with a directed communication topology. The proposed method extends the previous work in the context of passive system control and is suitable for more general directed communication graph topology. The communication delays and the controller saturation are also considered. Under the proposed theoretic framework, we show that the synchronization problems of networked Lagrange systems can be solved. As an illustration of our results, we study the synchronizing control for the networked manipulator systems. The simulation results show the effectiveness of the proposed algorithm.     2012 Vol. 38 (5): 882-888 [Abstract ] ( 36 ) [HTML 0KB] [PDF 603KB ]( 45 ) 　Authors 　Instruction 　FAQ 　Online Submission 　Manuscript Tracking 　Messages to Editors 　Online Review 　Editor-in-Chief 　Editorial Board 　Peer Review 　Volunteer to be a referee 　Reader 　Subscription 　Email Alert 　Member Service 　Download 　Template 　Copyright Agreement 　　　　　　　　　　　More... 　Links 　CNKI 　Wanfang 　　　　　　　　　　　More...              Copyright © 2006 Editorial office of Acta Automatica Sinica Tel: 010-62565763　E-mail: aas@mail.ia.ac.cn Supported by Beijing Magtech Co.Ltd Tel：010-62043307, E-mail: support@magtech.com.cn