Skip to main content
main-content

Über dieses Buch

The 2010 International Conference on Artificial Intelligence and Computational Intelligence (AICI 2010) was held October 23–24, 2010 in Sanya, China. The AICI 2010 received 1,216 submissions from 20 countries and regions. After rigorous reviews, 105 high-quality papers were selected for publication in the AICI 2010 proceedings. The acceptance rate was 8%. The aim of AICI 2010 was to bring together researchers working in many different areas of artificial intelligence and computational intelligence to foster the exchange of new ideas and promote international collaborations. In addition to the large number of submitted papers and invited sessions, there were several internationally well-known keynote speakers. On behalf of the Organizing Committee, we thank Hainan Province Institute of Computer and Qiongzhou University for its sponsorship and logistics support. We also thank the members of the Organizing Committee and the Program Committee for their hard work. We are very grateful to the keynote speakers, invited session organizers, session chairs, reviewers, and student helpers. Last but not least, we thank all the authors and participants for their great contributions that made this conference possible.

Inhaltsverzeichnis

Frontmatter

Applications of Artificial Intelligence

Application of RBF Neural Network in Short-Term Load Forecasting

A Radius Basic Function (RBF) neural network is proposed for the power load forecasting. RBF neural network can meet nonlinear recognition and process predition of the dynamic system, and has better adaptability to dynamic forecasting and prediction problem in mechnism. The RBF centres are determined by the orthogonal least squared (OLS) learning procedure. The effectiveness of the model and algorithm with the example of power load forecasting have been proved and approximation capability and learning speed of RBF neural network is better than BP neural network.

Yongchun Liang

Efficient Large Image Browser for Embedded Systems

Images obtained by digital cameras nowadays become larger and larger in size. The quality of the obtained images is very excellent, but on the other hand, they are hard to be displayed on such embedded systems as mobile phones, digital photo frames, etc., which have both very limited memory size and very small screen size. In this paper, an efficient large image browser for embedded systems is described. A set of approaches based on pixel resampling technology are proposed to make large images to be displayed effectively. The proposed browser consumes much less memory than many famous image browser softwares. Experimental evaluations of the proposed mechanism indicate that it is efficient and effective for large image display on embedded systems.

Yuanyuan Liu, Zhiwei He, Haibin Yu, Jinbiao Liu

Formalizing Ontology-Based Hierarchical Modeling Process of Physical World

The important step of model-based reasoning is to construct the abstraction model of physical world. In this paper, we introduce the concept of hierarchical ontology classes within the G-KRA model framework to realize the modeling knowledge sharing and reuse. Three kinds of ontology operators are defined and the formal process of them working is partly given. We also point out that constructing hierarchical ontology classes can be independent of the modeling process. Moreover we show how to automatically build up a hierarchical model of a world W based on the hierarchical ontology classes which is as well updated during the modeling. We expect that this work will rich the way to construct hierarchical model of physical world automatically and efficiently.

Nan Wang, Dantong OuYang, Shanwu Sun

Automated Problem Solving

Satisfiability Degree Analysis for Transition System

Classical model checking is not capable of solving the situation of uncertain systems or fault-tolerant systems which occur in the real world commonly. The technique of satisfiability degree (SD) for model checking is a efficient way to solve this problem. Finite paths transition system (FPTS) for calculating satisfiability degree of a LTL logic formula is given. Then, a more general situation about discrete-time Markov chains (DTMCs) is discussed. Then a case named leader election shows the practicability of satisfiability degree for transition system, which cannot be solved by classical model checking.

Yang Zhao, Guiming Luo

Research on Optimization Design of Bottom Angle of Drag-Reducing Structure on the Riblets Surface

Research on the drag-reducing structure of the haploid interval riblets surface is carried out with the numerical simulation, and in view of its flow field features, this paper appropriately deals with the computing domain, grids and flow parameters during the numerical calculation. By the adoption of iSIGHT software, structure parameters generated program of riblets surface, hydrodynamic numerical simulation software GAMBIT and FLUENT is integrated. With optimized design technology and the numerical simulation technology, the optimization design of the bottom angle on the riblets surface is made by adopting the two-stage combining optimization method which includes overall search based on adaptive simulated annealing algorithm and partial optimal search of sequential quadratic programming. The results show that after the optimization design, the relative amount of drag reduction of the riblets surface reaches 3.506 %. Compared with the initial value, the relative amount of drag reduction increases and the drag-reducing effect is remarkable. This provides a new direction for the research on drag reduction of riblets surface.

Qi-feng Zhu, Bao-wei Song, Peng Wang

Towards Analysis of Semi-Markov Decision Processes

We investigate Semi-Markov Decision Processes (SMDPs). Two problems are studied, namely, the time-bounded reachability problem and the long-run average fraction of time problem. The former aims to compute the maximal (or minimum) probability to reach a certain set of states within a given time bound. We obtain a Bellman equation to characterize the maximal time-bounded reachability probability, and suggest two approaches to solve it based on discretization and randomized techniques respectively. The latter aims to compute the maximal (or minimum) average amount of time spent in a given set of states during the long run. We exploit a graph-theoretic decomposition of the given SMDP based on maximal end components and reduce it to linear programming problems.

Taolue Chen, Jian Lu

Automatic Programming

Stability of Equilibrium Solution and Periodical Solution to Cohen-Grossberg Neural Networks

In this paper, we study delayed reaction-diffusion Cohen-Grossberg neural networks with Dirichlet boundary conditions. By using topology degree theory and constructing suitable Lyapunov functional, some sufficient conditions are given to ensure the existence, uniqueness and globally exponential stability of the equilibrium point. At the same time, another sufficient conditions are also given to ensure the existence and exponential convergence of the periodical solution.

Jingsheng Lei, Ping Yan, Teng Lv

Exponential Synchronization of Delayed Fuzzy Cohen-Grossberg Neural Networks with Reaction Diffusion Term

In this paper, the exponential synchronization of delayed fuzzy Cohen-Grossberg neural networks is discussed. Some sufficient conditions are given to ensure the exponential synchronization of the drive-response delayed fuzzy cellular neural networks.

Teng Lv, Ping Yan

Magnetic Field Extrapolation Based on Improved Back Propagation Neural Network

Magnetic anomaly created by ferromagnetic ships may make them vulnerable to detections and mines. In order to reduce the anomaly, it is important to evaluate magnetic field firstly. Underwater field can be measured easily, but upper air field is hard to be got. To achieve it, a model able to predict upper air magnetic field from underwater measurements is required. In this paper, a Back Propagation (BP) model has been built and it can escape from local optimum thanks to optimizing the initial weights and threshold values by Particle Swarm Optimization (PSO) algorithm. The method can avoid many problems from linear model and its high accuracy and good robustness have been tested by a mockup experiment.

Li-ting Lian, Chang-han Xiao, Sheng-dao Liu, Guo-hua Zhou, Ming-ming Yang

Sparse Deep Belief Net for Handwritten Digits Classification

It has been shown that the Deep Belief Network is good at modeling input distribution, and can be trained efficiently by the greedy layer-wise unsupervised learning. Hoglak Lee et al. (2008) introduced a sparse variant of the Deep Belief Network, which applied the Gaussian linear units to model the input data with a sparsity constraint. However, it takes much more weight updates to train the RBM (Restricted Boltzmann Machine) with Gaussian visible units, and the reconstruction error is much larger than training an RBM with binary visible units. Here, we propose another version of Sparse Deep Belief Net which applies the differentiable sparse coding method to train the first level of the deep network, and then train the higher layers with RBM .This hybrid model, combining the advantage of the Deep architecture and the sparse coding model, leads to state-of-the-art performance on the classification of handwritten digits.

Jiongyun Xie, Hongtao Lu, Deng Nan, Cai Nengbin

Data Mining and Knowledge Discovering

Multisensor Image Fusion Using a Pulse Coupled Neural Network

Multisensor image fusion has its effective utilization for surveillance. In this paper, we utilize a pulse coupled neural network method to merge images from different sensors, in order to enhance visualization for surveillance. On the basis of standard mathematical model of pulse coupled neural network, a novel step function is adopted to generate pulses. Subjective and objective image fusion performance measures are introduced to assess the performance of image fusion schemes. Experimental results show that the image fusion method using pulse coupled neural network is effective to merge images from different sensors.

Yi Zheng, Ping Zheng

Real-Time Performance Reliability Assessment Method Based on Dynamic Probability Model

Most previous reliability estimation methods are researched on the assumption of empirical information or prior distribution which is difficult to be acquired in practice. To solve this problem, a real-time reliability assessment method based on Dynamic Probability Model is proposed. The primary step is to establish a Dynamic Probability Model on the basis of nonparametric Parzen window estimating method, and the sliding time-window technique is used to pick statistical samples respectively, then conditional probability density of performance degradation data is estimated. A sequential probability density curve is used to trace the performance degradation process, and probability distribution function on performance degradation data which exceeds the failure threshold is regarded as reliability indicator. Meanwhile, the failure rate is calculated. By analyzing the data from high pressure water descaling pump in the process of failure, it is verified that this method contributes individual equipment to estimate reliability with inadequate empirical information.

Cheng Hua, Qing Zhang, Guanghua Xu, Jun Xie

Decomposing Data Mining by a Process-Oriented Execution Plan

Data mining deals with the extraction of hidden knowledge from large amounts of data. Nowadays, coarse-grained data mining modules are used. This traditional black box approach focuses on specific algorithm improvements and is not flexible enough to be used for more general optimization and beneficial component reuse. The work presented in this paper elaborates on decomposing data mining tasks as data mining execution process plans which are composed of finer-grained data mining operators. The cost of an operator can be analyzed and provides means for more holistic optimizations. This process-based data mining concept is evaluated via an OGSA-DAI based implementations for association rule mining which show the feasibility of our approach as well as the re-usability of some of the data mining operators.

Yan Zhang, Honghui Li, Alexander Wöhrer, Peter Brezany, Gang Dai

An Efficient Distributed Subgraph Mining Algorithm in Extreme Large Graphs

Graph mining plays an important part in the researches of data mining, and it is widely used in biology, physics, telecommunications and Internet in recently emerging network science. Subgraph mining is a main task in this area, and it has attracted much interest. However, with the growth of graph datasets, most of these former works which mainly rely on single chip computational capacity, cannot process massive graphs. In this paper, we propose a distributed method in solving subgraph mining problems with the help of MapReduce, which is an efficient method of computing. The candidate subgraphs are reduced efficiently according to the degrees of nodes in graphs. The results of our research show that the algorithm is efficient and scalable, and it is a better solution of subgraph mining in extreme large graphs.

Bin Wu, YunLong Bai

Spatio-Temporal Clustering of Road Network Data

This paper addresses spatio-temporal clustering of network data where the geometry and structure of the network is assumed to be static but heterogeneous due to the density of links varies cross the network. Road network, telecommunication network and internet are of these type networks. The thematic properties associated with the links of the network are dynamic, such as the flow, speed and journey time are varying in the peak and off-peak hours of a day. Analyzing the patterns of network data in space-time can help the understanding of the complexity of the networks Here a spatio-temporal clustering (STC) algorithm is developed to capture such dynamic patterns by fully exploiting the network characteristics in spatial, temporal and thematic domains. The proposed STC algorithm is tested on a part of London’s traffic network to investigate how the clusters overlap on different days.

Tao Cheng, Berk Anbaroglu

A Sampling Based Algorithm for Finding Association Rules from Uncertain Data

Since there are many real-life situations in which people are uncertain about the content of transactions, association rule mining with uncertain data is in demand. Most of these studies focus on the improvement of classical algorithms for frequent itemsets mining. To obtain a tradeoff between the accuracy and computation time, in this paper we introduces an efficient algorithm for finding association rules from uncertain data with sampling-SARMUT, which is based on the FAST algorithm introduced by Chen et al. Unlike FAST, SARMUT is designed for uncertain data mining. In response to the special characteristics of uncertainty, we propose a new definition of ”distance” as a measure to pick representative transactions. To evaluate its performance and accuracy, a comparison against the natural extension of FAST is performed using synthetic datasets. The experimental results show that the proposed sampling algorithm SARMUT outperforms FAST algorithm, and achieves up to 97% accuracy in some cases.

Zhu Qian, Pan Donghua, Yang Guangfei

Distributed AI and Agents

Multi-agent System Collaboration Based on the Relation-Web Model

In a multi-agent system, the agents have the capability to work collaboratively to solve the complex task. In recent years social computing provides a new perspective for multi-agent collaboration. This paper first introduces the role model and role relationships including role inheritance, role preference, role binding etc. Then a relation-web model is proposed referring to the social computing research. To a great extent, the relation-web model is used to simulate the social collaboration. The relation weight called trust degree is updated according to their collaboration result. When a complex task is assigned to some agents, the agents will construct the relation-webs for the sub-tasks completion. Finally, to test the relation-web model, an experiment is designed to predict the electricity consumption. The result proves the model to be available and useful while simulating the multi-agent collaboration process for solving the practical problem.

Maoguang Wang, Hong Mei, Wenpin Jiao, Junjing Jie, Tingxun Shi

Research on A Novel Multi-Agents Dynamic Cooperation Method Based on Associated Intent

With the information explosion speeds up the increasing of computing complexity rapidly, the traditional centralized computing patterns are under great pressure to process those large-scale distributed information. However, the agent-based computation and high-level interaction protocols foster the modern computation and distributed information processing successfully. The multi-agent system (MAS) plays an important role in the analysis of the humaninteraction theory and model building. This study focuses on the formal description of MAS, the conflict-resolving mechanisms and the negotiation in MAS. The communication between agents has some special requirements. One of them is asynchronous communication. Used communication sequence process (CSP) to descript a model of agents communication with shared buffer channel. The essence of this model is very suitable for the multiagents communication, so it is a base for our next step job. Based on the communication model, explored the distributed tasks dealing method among joint intention agents and with description of relation between tasks we give a figure of agents’ organization. Agents communicate with each other in this kind of organization. The semantics of agent communication is another emphasis in this paper. With the detailed description of agents’ communication process, given a general agent automated negotiation protocol based on speech act theory in MAS, then we use CSP to verify this protocol has properties of safety and liveness, so prove it is logic right. At last a frame of this protocol’s realization was given.

Weijin Jiang, Xiaoling Ding, Yuhui Xu, Wang Chen, Wei Chen

Expert and Decision Support Systems

Automobile Exhaust Gas Detection Based on Fuzzy Temperature Compensation System

A temperature compensation scheme of detecting automobile exhaust gas based on fuzzy logic inference is presented in this paper. The principles of the infrared automobile exhaust gas analyzer and the influence of the environmental temperature on analyzer are discussed. A fuzzy inference system is designed to improve the measurement accuracy of the measurement equipment by reducing the measurement errors caused by environmental temperature. The case studies demonstrate the effectiveness of the proposed method. The fuzzy compensation scheme is promising as demonstrated by the simulation results in this paper.

Zhiyong Wang, Hao Ding, Fufei Hao, Zhaoxia Wang, Zhen Sun, Shujin Li

Technical Efficiency Analysis of New Energy Listed Companies Based on DEA

This paper uses the DEA model to analyses the technical efficiency of representative listed companies in China new energy industry representative.Through the analysis and solution for the data on three consecutive years, the results in a certain extent, reflect the development and business efficiency of new energy industry, and provide relevant decision-making information on technical efficiency for new energy industry operation, with strong reliability and practicality.

Gao Chong, Zhang Jian-ze, Li Xiao-dong

Fuzzy Logic and Soft Computing

Research on Differential Evolution Algorithm Based on Interaction Balance Method for Large-Scale Industrial Processes of Fuzzy Model

As a result of slow perturbations, the mathematical model of an actual system is difficult to be accurate. So when optimizing large-scale industrial process, the mathematical model and the actual system does not match, that is model-actual difference. Large-scale industrial process optimization based on fuzzy model is an effective way of this issue. However, the optimization model is the process of establishing a non-linear programming model. So, differential evolution algorithm is studied this paper to solve the problems of large-scale industrial processes optimization based on fuzzy models. To solve model-actual differences is mainly to solve fuzzy nonlinear problems: firstly, differential evolution algorithm is studied for solving fuzzy nonlinear problems in this paper. Then the combination of fuzzy nonlinear problems and the interaction balance method coordinated approach of large-scale industrial processes is proposed. Last, simulation results show the validity of the method which this paper studies.

He Dakuo, Zhao Yuanyuan, Wang Lifeng, Chang Hongrui

Fuzzy Control System Design for Solar Energy with PCM Tank to Fresh Air Conditioning

With the help of Solar-collector and PCM (Phase Change Material) tank, an experimental system has been set up. In this system, the framework of the system is based on fuzzy algorithm. To provide a better precision of the fresh-air temperature, the system works in three different modes which need three different control rules. In this paper, these three modes are introduced firstly, and then a design of the three fuzzy controllers is described in detail. Meanwhile, the automatic transition process of the three modes in different conditions is studied. Finally, through the analysis of experimental data, it is proved that the system can provide stable fresh air temperature according to different setting temperature.

Jun Yang, Ailin Xiao, Lili Wang

Intelligent Information Fusion

An Efficient Method for Target Extraction of Infrared Images

This paper proposes an efficient method to extract targets from an infrared image. First, the regions of interests (ROIs) which contain the entire targets and a little background region are detected based on the variance weighted information entropy feature. Second, the infrared image is modeled by Gaussian Markov random field, and the ROIs are used as the target regions while the remaining region as the background to perform the initial segmentation. Finally, by searching solution space within the ROIs, the targets are accurately extracted by energy minimization using the iterated condition mode. Because the iterated segmentation results are updated within the ROIs only, this coarse-to-fine extraction method can greatly accelerate the convergence speed and efficiently reduce the interference of background noise. Experimental results of the real infrared images demonstrate that the proposed method can extract single and multiple infrared objects accurately and rapidly.

Ying Li, Xingjin Mao

Second Order Central Difference Filtering Algorithm for SINS/GPS Integrated Navigation in Guided Munitions

With the SINS/GPS integrated navigation system, the guided munitions can be carried out in complex weather, and have the high positioning navigation accuracy. This paper deduces key matrix of the second order central difference filtering (CDF2) equation. The CDF2 is described as a sigma point filter in a unified way where the filter linearizes the nonlinear dynamic and measurement functions by using an interpolation formula through systematically chosen sigma points. The effect which the key parameters of CDF2 bring to information fusion is analyzed qualitatively. The structure of loose integration is also given. According to the test data, the fusion algorithm based on CDF2 is applied. Compared to the original algorithm in longitude, latitude, altitude and velocity, the orientation precision is improved greatly.

Lei Cai, Xuexia Zhang

Adaptive Signal Processing for ARX System Disturbed by Complex Noise

The inverse of the Fisher information matrix can be decided by the system input sequence and the disturbance variance if a Gaussian noise is involved. The lower bound mean-square error matrix of any unbiased estimator is given by Cramer-Rao Lemma. When a system is disturbed by some biased noises, the classical Fisher information matrix would be not valid. The bound is not fitted when a biased estimator is implemented. Signal processing for ARX model disturbed by complex noise is concerned in this paper. Cramer-Rao bound of a biased estimation is obtained. An adaptive signal processing algorithm for identification of ARX system disturbed by biased estimation is proposed. Some experiments are included to verify the efficiency of the new algorithm.

Yulai Zhang, Guiming Luo

Abstraction for Model Checking the Probabilistic Temporal Logic of Knowledge

Probabilistic temporal logics of knowledge have been used to specify multi-agent systems. In this paper, we introduce a probabilistic temporal logic of knowledge called PTLK for expressing time, knowledge, and probability in multi-agent systems. Then, in order to overcome the state explosion in model checking PTLK we propose an abstraction procedure for model checking PTLK. The rough idea of the abstraction approach is to partition the state space into several equivalence classes which consist of the set of abstract states. The probability distribution between abstract states is defined as an interval for computing the approximation of the concrete system. Finally, the model checking algorithm in PTLK is developed.

Conghua Zhou, Bo Sun, Zhifeng Liu

Intelligent Scheduling

Research on Double-Objective Optimal Scheduling Algorithm for Dual Resource Constrained Job Shop

To solve the double-objective optimization of dual resource constrained job shop scheduling, an inherited genetic algorithm is proposed. In the algorithm, evolutionary experience of parent population is inherited by the means of branch population supplement based on pheromones to accelerate the convergence rate. Meanwhile, the activable decoding algorithm based on comparison among time windows, the resource crossover operator and resource mutation operator, which are all established based on four-dimensional coding method are utilized with reference to the character of dual resource constrained to improve the overall searching ability. Furthermore, the championship selection strategy based on Pareto index weakens the impact of the Pareto level of chromosomes obviously. The elitist preservation strategy guarantees reliable convergence of the algorithm. Simulation results show that the performance of the proposed inherited GA is effective and efficient.

Li Jingyao, Sun Shudong, Huang Yuan, Niu Ganggang

MDA Compatible Knowledge– Based IS Engineering Approach

Enhancement of MDA process with Knowledge Base Subsystem is aimed to reduce risk of project failures caused by inconsistent user requirements caused by insufficient problem domain knowledge. The enhancement of IS development environment with Enterprise Knowledge Base is discussed in this article. The major parts of Knowledge Base Subsystem are Enterprise Meta-Model, Enterprise Model and transformation algorithms.

Audrius Lopata, Martas Ambraziunas

Intelligent Signal Processing

ARIMA Signals Processing of Information Fusion on the Chrysanthemum

Weak electrical signals of the plant were tested by a touching test system of self-made double shields with platinum sensors. Tested data of electrical signals were denoised with the wavelet soft threshold. A novel autoregressive integrated moving average (ARIMA) model of weak electric signals of the chrysanthemum was constructed by the information fusion technology for the first time, that is, Xt =1.93731Xt-1 - 0.93731Xt-2 +

ε

t + 0.19287

ε

t–1 - 0.4173 Xt-2 - 0.17443 Xt-4 - 0.07764 Xt-5 - 0.06222 Xt-7. A fitting standard deviation was 1.814296. It has a well effect that the fitting variance and standard deviation of the model are the minimum. It is very importance that the plant electric signal with the data fusion is to understand self-adapting regulations on the growth relationship between the plant and environments. The forecast data can be used as preferences for the intelligent system based on the adaptive characters of plants.

Lanzhou Wang, Qiao Li

Noise Uncertainty Study of the Low SNR Energy Detector in Cognitive Radio

It’s well known that a signal-noise-ratio(SNR) threshold named ”SNR wall” [5] appears in the energy detection due to the noise uncertainty. It makes energy detector(ED) highly non-robust in low SNR environment. In this paper, a log-normal distribution model of the noise power uncertainty is proposed. The detection performances of the energy detector are analyzed based on the proposed noise model. The numerical results illustrate that not only local but also cooperative low SNR Eds are reduced to invalid for a biggish noise power uncertainty.

Guoqing Ji, Hongbo Zhu

Machine Learning

A Realtime Human-Computer Ensemble System: Formal Representation and Experiments for Expressive Performance

A human-computer ensemble system is one of time concerned cooperative systems, which performs secondo of an ensemble played by a computer-controlled piano cooperating with primo played by a human performer. For creating expressive performance, a rehearsal program is adopted to this system. By the rehearsal program, the system learns the tendency of the expression that the human performer thinks and/or plans. To do so, the program records his/her performance of solo and calculates the tendency, which depends on not only the performer but also the composition of music itself. Hence, it is necessary for the expressive performance to analyze the score of the composition and to experiment dependent on it.

This system is an example of intelligent realtime programs appropriate for formal verification and analysis.

N

Σ-labeled calculus is a formal system for verification for such time-concerned programs.

In this paper, a logical specification and experimental results for an expressive ensemble system will be introduced.

Tetsuya Mizutani, Shigeru Igarashi, Tatsuo Suzuki, Yasuwo Ikeda, Masayuki Shio

A New Smooth Support Vector Machine

A new Smooth Support Vector Machine (SSVM) is proposed and is called NSSVM for short. Different from traditional SSVM that treats perturbation formulation of SVM, NSSVM treats standard 2-norm error soft margin SVM. Different from traditional SSVM that uses the 2-norm of the Lagrangian multipliers vector to roughly substitute that of the weight of the separating hyperplane, which makes the obtained smooth model unequal to the primal program; NSSVM takes into account the connotative relation between the primal and dual program to transform the original program to a new smooth one. Numerical experiments on several UCI datasets demonstrate that NSSVM has higher precisions than existing methods.

Jinjin Liang, De Wu

Convergence of GCM and Its Application to Face Recognition

We mainly generalize consistency method in semi-supervised learning by expanding kernel matrix,denoted by GCM(Generalized Consistency Method), and study its convergence. Aimed at GCM,we give the detailed proof for condition of convergence. Moreover,we further study the validity of some variants of GCM. Finally we conduct the experimental study on the parameters involved in GCM to face recognition. Meanwhile, the performance of GCM and its some variants are compared with that of support vector machine methods.

Kai Li, Xinyong Chen, Nan Yang, Xiuchen Ye

Designing a Multi-label Kernel Machine with Two-Objective Optimization

In multi-label classification problems, some samples belong to multiple classes simultaneously and thus the classes are not mutually exclusive. How to characterize this kind of correlations between labels has been a key issue for designing a new multi-label classification approach. In this paper, we define two objective functions, i.e., the number of relevant and irrelevant label pairs which are ranked incorrectly, and the model regularization term, which depict the correlations between labels and the model complexity respectively. Then a new kernel machine for multi-label classification is constructed using two-objective minimization and solved by fast and elitist multi-objective genetic algorithm, i.e., NSGA-II. Experiments on the benchmark data set Yeast illustrate that our multi-label method is a competitive candidate for multi-label classification, compared with several state-of-the-art methods.

Hua Xu, Jianhua Xu

Collision Detection Algorithm in Virtual Environment of Robot Workcell

Rapid and accurate collision detection is necessary to robot programming and simulation system based on virtual reality. A collision detection algorithm based on the data structure of the pre-built scene graph is proposed. It is the hierarchical accurate method. The architecture of collision detection system is established according to the hierarchy of scene graph for robot workcell, and implementation process of the collision detection algorithm is given. The collision detection system includes four collision detection managers. Groups filter manager, objects filter manager and faces & objects intersection manager are used in the broad phase of the collision detection, and they improve the speed of the algorithm by AABB intersection test and the layer by layer filtration. Polygons intersection manager is used in the narrow phase, and it ensures the accuracy of the algorithm. The result of the 3D graphics simulation experiment proves the effectiveness and feasibility of the way introduced.

Qian Ren, Dongmei Wu, Shuguo Wang, Yili Fu, Hegao Cai

Machine Vision

A First Step towards Hybrid Visual Servoing Control Based on Image Moments

The paper is concerned with a specific class of visual servoing problem, in which camera motion (including both translation and rotation) are constraint to be in the Z-axis direction. Such a constraint condition makes it possible to find appropriate image moments reflecting object depth and orientation. Image moments, as a kind of global image features, can be benefit to the performance of visual servoing system, such as insensitivity to image noise, nonsingularity in image Jacobian, and etc. In the paper, the mathematic relationships between image moments and object-depth-and-orientation are firstly introduced. Then appropriate image moments are selected, on the basis of which a hybrid visual servoing system is build. In our system, visual servoing controller consists of two parts: one is called translation controller which is in charge of object depth control, the other is called rotation controller which controls object orientation. The simulation results show that our hybrid visual servoing system performances well with a high accuracy.

Xiaojing Shen, Dongmei Huang, Xiaoxia Qin

A Novel Motion Detection Approach for Large FOV Cameras

Moving objects detection by an also moving camera plays an important role in driver assistance systems and robot navigations. Many motion detection methods have been proposed until now. But most of them are based on normal cameras with a limited view and not suitable for large FOV (field of view) cameras. For motion detection using large FOV cameras, there are two main challenges. One comes from difficulties to tell moving objects from moving background due to camera motion. The other comes from the image distortion brought by large FOV cameras. These two problems are solved in our approach by a novel motion detector which can be considered as a special motion constraint based on virtual planes. The experimental results under various scenes illustrate the effectiveness of this work.

Hongfei Yu, Wei Liu, Bobo Duan, Huai Yuan, Hong Zhao

Large Scale Visual Classification via Learned Dictionaries and Sparse Representation

We address the large scale visual classification problem. The approach is based on sparse and redundant representations over trained dictionaries. The proposed algorithm firstly trains dictionaries using the images of every visual category, one category has one dictionary. In this paper, we choose the K-SVD algorithm to train the visual category dictionary. Given a set of training images from a category, the K-SVD algorithm seeks the dictionary that leads to the best representation for each image in this set, under strict sparsity constraints. For testing images, the traditional classification method under the large scale condition is the

k

-nearest-neighbor method. And in our method, the category result is through the reconstruction residual using different dictionaries. To get the most effective dictionaries, we explore the large scale image database from the Internet [2] and design experiments on a nearly 1.6 million tiny images on the middle semantic level defined based on WordNet. We compare the image classification performance under different image resolutions and

k

-nearest-neighbor parameters. The experimental results demonstrate that the proposed algorithm outperforms

k

-nearest-neighbor in two aspects: 1) the discriminative capability for large scale visual classification task, and 2) the average running time of classifying one image.

Zhenyong Fu, Hongtao Lu, Nan Deng, Nengbin Cai

Semi-supervised Nearest Neighbor Discriminant Analysis Using Local Mean for Face Recognition

Feature extraction is the key problem of face recognition. In this paper, we propose a new feature extraction method, called semi-supervised local mean-based discriminant analysis (SLMNND). SLMNND aims to find a set of projection vectors which respect the discriminant structure inferred from the labeled data points, as well as the intrinsic geometrical structure inferred from both labeled and unlabeled data points. Experiments on the famous ORL and AR face image databases demonstrate the effectiveness of our method.

Caikou Chen, Pu Huang, Jingyu Yang

Multi-agent Systems

Decentralized Cohesive Motion Control of Multi-agent Formation in 3-Dimensional Space

In this paper, we generalize a set of decentralized control laws for cohesive motion of 3-dimensional multi-agent formation based on point-agent system model, which has been originally introduced for 2-dimensional formation. According to the numbers of degree of freedom, we investigate the persistency of multi-agent formation system. Then we design the corresponding control laws for the different class agents in the multi-agent formation. As a result, the 3-dimensional persistent formation can move cohesively along the trajectory as the target of the formation moves continuously along the prescribed trajectory. The effectiveness of the proposed control laws is demonstrated via simulation results.

Ran Zhao, Yongguang Yu, Guoguang Wen

A Model for Cooperative Design Based on Multi-Agent System

For lack of information exchange and mechanism to share knowledge,the traditional distributed collaborative design mode is more difficult to adapt to the increasing development needs of product design. This paper propose an collaborative design mode based on multi-agent, according to the environment of the collaborative design and design goal, the whole process of product development is considered by this model effectively, so it will shorten the time of design, reduce the cost of product, improve quality and response the need of customers rapidly.

Hua Chen, Jun Zhao, Bo Sun

Natural Language Processing

Scaling Up the Accuracy of Bayesian Classifier Based on Frequent Itemsets by M-estimate

Frequent Itemsets Mining Classifier (FISC) is an improved Bayesian classifier which averaging all classifiers built by frequent itemsets. Considering that in learning Bayesian network classifier, estimating probabilities from a given set of training examples is crucial, and it has been proved that m-estimate can scale up the accuracy of many Bayesian classifiers. Thus, a natural question is whether FISC with m-estimate can perform even better. Response to this problem, in this paper, we aim to scale up the accuracy of FISC by m-estimate and propose new probability estimation formulas. The experimental results show that the Laplace estimate used in the original FISC performs not very well and our m-estimate can greatly scale up the accuracy, it even outperforms other outstanding Bayesian classifiers used to compare.

Jing Duan, Zhengkui Lin, Weiguo Yi, Mingyu Lu

Aggressive Dimensionality Reduction with Reinforcement Local Feature Selection for Text Categorization

A major problem of text categorization is the high dimensionality of the input feature space. This paper proposes a novel approach for aggressive dimensionality reduction in text categorization. This method utilizes the local feature selection to obtain more positive terms and then scales the weighting in the global level to suit the classifier. After that the weighting is enhanced with the feature selection measure to improve the distinguishing capability. The validity of this method is tested on two benchmark corpuses by the SVM classifier with four standard feature selection measures.

Wenbin Zheng, Yuntao Qian

Neural Networks

3D Shape Representation Using Gaussian Curvature Co-occurrence Matrix

Co-occurrence matrix is traditionally used for the representation of texture information. In this paper, the co-occurrence matrix is combined with Gaussian curvature for 3D shape representation and a novel 3D shape description approach named Gaussian curvature co-occurrence matrix is proposed. Normalization process to Gaussian curvature co-occurrence matrix and the invariants independence of the translation, scaling and rotation transforms are demonstrated. Experiments indicate a better classification rate and running complexity to objects with slight different shape characteristic compared with traditional methods.

Kehua Guo

Nonlinear System Identification of Bicycle Robot Based on Adaptive Neural Fuzzy Inference System

System identification is the basis of designing control system. The bicycle robot is an under-actuated, non-linear, non-integrated system with lateral instability, it’s two wheels are longitudinal and has non-sliding contact with the ground, meanwhile it’s dynamic characteristics are complicated. So it is very difficult to set up more precise dynamics model. While precise model of complex system often requires more complex control design and calculation. In this paper, linear ARX model and nonlinear ANFIS model are proposed. The identifications of bicycle robot system are completed through the data of handlebar angle and those of inclination angle which are gathered when bicycle robot is stable. Simulation result by ANFIS based on T-S model could be very similar to the actual test data of bicycle robot sysytem, and it’s identification precision is higher than that of linear ARX model. The obtained conclusions of fuzzy inference between input and output by above identificaton methods can provide some reference value for effective control on bicycle robot system in future.

Xiuli Yu, Shimin Wei, Lei Guo

Transmission: A New Feature for Computer Vision Based Smoke Detection

A novel and effective approach is proposed in this paper to detect smoke using transmission from image or video frame. Inspired by the airlight-albedo ambiguity model, we introduce the concept of transmission as a new essential feature of smoke, which is employed to detect the smoke and also determine its corresponding thickness distribution. First, we define an optical model for smoke based on the airlight-albedo ambiguity model. Second, we estimate the preliminary smoke transmission using dark channel prior and then refine the result through soft matting algorithm. Finally, we use transmission to detect smoke region by thresholding and obtain detailed information about the distribution of smoke thickness through mapping transmissions of the smoke region into a gray image. Our method has been tested on real images with smoke. Compared with the existing methods, experimental results have proved the better efficiency of transmission in smoke detection.

Chengjiang Long, Jianhui Zhao, Shizhong Han, Lu Xiong, Zhiyong Yuan, Jing Huang, Weiwei Gao

A Novel Features Design Method for Cat Head Detection

In this paper we have proposed a new novel features model whichdesigned to robustly detect the highly variable cat head patterns.Do not like human, cats usually have distinct different face, pose,appearance and different scales of ears, eyes and mouth. So manysignificant features on human face detection have presented but itis not satisfying to use them on cat head. We have designed a newfeatures model by ideally combining the histogram frame withGLCM-based (gray level co-occurrence matrix) texture features todescribe both the shape information of cat’s head and textureinformation of cat’s eyes, ears and mouth in detail. SVM-basedclassifier achieves the detection results. Extensive experimentalresults illustrating the high detection rate with low false alarm.

Hua Bo

Time Serial Model of Rock Burst Based on Evolutionary Neural Network

Rock burst is a mining disaster, which can bring large harm. The theory studies have showed that, rock burst is a kind of dynamic phenomenon of mining rock, and is a kind of dynamic disaster from mining. Considering the dynamic character of rock burst, the construction time serial model of rock burst is studied by evolutionary neural network based on immunized evolutionary programming. At last, the proposed model is tested by a real magnitude series of a mine. The result has showed that, evolutionary neural network model not only has high approaching precision, but also has high predicting precision, and is a good method to construct the non-linear model of rock burst.

Wei Gao

Multilayer Perceptron Network with Modified Sigmoid Activation Functions

Models in today’s microcontrollers, e.g. engine control units, are realized with a multitude of characteristic curves and look-up tables. The increasing complexity of these models causes an exponential growth of the required calibration memory. Hence, neural networks, e.g. multilayer perceptron networks (MLP), which provide a solution for this problem, become more important for modeling. Usually sigmoid functions are used as membership functions. The calculation of the therefore necessary exponential function is very demanding on low performance microcontrollers. Thus in this paper a modified activation function for the efficient implementation of MLP networks is proposed. Their advantages compared to standard look-up tables are illustrated by the application of an intake manifold model of a combustion engine.

Tobias Ebert, Oliver Bänfer, Oliver Nelles

Pattern Recognition

Kernel Oblique Subspace Projection Approach for Target Detection in Hyperspectral Imagery

In this paper, a kernel-based nonlinear version of the oblique subspace projection (OBSP) operator is defined in terms of kernel functions. Input data are implicitly mapped into a high-dimensional kernel feature space by a nonlinear mapping, which is associated with a kernel function. The OBSP expression is then derived in the feature space, which is kernelized in terms of the kernel functions in order to avoid explicit computation in the high-dimensional feature space. The resulting kernelized OBSP algorithm is equivalent to a nonlinear OBSP in the original input space. Experimental results based on simulated hyperspectral data and real hyperspectral imagery shows that the kernel oblique subspace projection (KOBSP) outperforms the conventional OBSP.

Liaoying Zhao, Yinhe Shen, Xiaorun Li

Text-Independent Voice Conversion Based on Kernel Eigenvoice

Almost of the current spectral conversion methods required parallel corpus containing the same utterances from source and target speakers, which was often inconvenient and sometimes hard to fulfill. This paper proposed a novel algorithm for text-independent voice conversion, which can relax the parallel constraint. The proposed algorithm was based on speaker adaptation technique of kernel eigenvoice, adapting the conversion parameters derived for the pre-stored pairs of speakers to a desired pair, for which only a nonparallel corpus was available. Objective evaluation results demonstrated that the proposed kernel eigenvoice algorithm can effectively improve converted spectral similarity in a text-independent manner.

Yanping Li, Linghua Zhang, Hui Ding

An Effective Method for SAR Automatic Target Recognition

Since synthetic aperture radar (SAR) images are very sensitive to the pose variation of targets, SAR automatic target recognition (ATR) is a well-known very challenging problem. This paper introduces an effective method for SAR ATR by using a combination of kernel singular value decomposition (KSVD) and principal component analysis (PCA) for feature extraction and the nearest neighbor classifier (NNC) for classification. Experiments are carried out on the Moving and Stationary Target Acquisition and Recognition (MSTAR) public database to evaluate the performance of the proposed method in comparison with the traditional PCA, singular value decomposition (SVD), kernel PCA (KPCA) and KSVD. The results demonstrate that the proposed method performs much better than the other methods with a right recognition rate up to 95.75%.

Ying Li, Hongli Gong

Semi-supervised Learning by Spectral Mapping with Label Information

A novel version of spectral mapping for partially labeled sample classification is proposed in this paper. This new method adds the label information into the mapping process, and adopts the geodesic distance rather than Euclidean distance as the measure of the difference between two data points. The experimental results show that the proposed method yields significant benefits for partially labeled classification with respect to the previous methods.

Zhong-Qiu Zhao, Jun Gao, Xindong Wu

Drop Fingerprint Recognition Based on Self-Organizing Feature Map

Drop analysis technology developed rapidly, the recognition of drop fingerprint become more and more important. It discussed about drop analysis technology and the methods to recognize liquid drop fingerprint. With the self-learning, self-organizing and out-supervision, self-organizing feature map network is suitable to use in drop fingerprint recognition. By MATLAB simulation, a SOM neural network which has been trained is established. Two groups of samples are identified. The identification ratio of one group is 97.5 percent, and the other group is 95 percent. The recognition performance achieved the goal as expected.

Jie Li, Qing Song, Yuan Luo, Cunwei Zou

Nonlinear Complementarity Problem and Solution Methods

This paper provides a survey to some of recent developments in the field of nonlinear complementarity problems (NCP). Some existence conditions of solution to the NCP are given according to the monotonicity of the functions, and corresponding NCP examples are demonstrated respectively. Meanwhile, a couple of different solution methods for NCP are described. Finally, we provide a brief summary of current research trends.

Longquan Yong

Robotics

A New Path Planning Method for a Shape-Shifting Robot

A shape-shifting robot with changeable configurations can accomplish search and rescue tasks which could not be achieved by manpower sometimes. The accessibility of this robot to uneven environment was efficiently enlarged by changing its configuration. In this paper, a path planning method is presented that integrates the reconfigurable ability of the robot with the potential field law. An adaptive genetic algorithm is applied to solve effectively the local minimum problem. The experiments show that the robot’s configurations can be changed to perform the path planning with the environmental variation. Moreover, the path has been shortened effectively.

Mengxin Li, Ying Zhang, TongLin Liu, Chengdong Wu

Pedestrian Gait Classification Based on Hidden Markov Models

Analysis of human activity from video sequences is one of the hottest and difficult research areas in computer visions. Because of the fact that human continuous motion can be decomposed into an image sequence based on time, state space method is applied in this paper. First, Silhouettes are extracted using the Background Subtraction method and features are represented by moment. Then a method using recursion method for establishment of the standard gait state sequence is proposed. In order to determine whether the behavior is abnormal in different scenarios, wavelet moment is used to extract features of the human body images, and then recognizes the moving human bodies activity based on Discrete Hidden Markov Model. The experiment tests show some encouraging results also indicates the algorithm has very small leak-examining and mistake-examining-rate which indicate that the method could be a choice for solving the problem but more tests are required.

Weihua Wang, Zhijing Liu

L-Infinity Norm Minimization in the Multiview Triangulation

Triangulation is an important part of numerous computer vision systems. The multiview triangulation problem is often solved by minimizing a cost function that combines the reprojection errors in the 2D images. In this paper, we show how to recast multiview triangulation as quasi-convex optimization under the L-infinity norm. It is shown that the L-infinity norm cost function is significantly simpler than the L2 cost. In particular L-infinity norm minimization involves finding the minimum of a cost function with a single global minimum on a convex parameter domain. These problems can be efficiently solved using second-order cone programming. We carried out experiment with real data to show that L-infinity norm minimization provides a more accurate estimate and superior to previous approaches.

Yang Min

Erratum

Erratum to: An Efficient Method for Target Extraction of Infrared Images

The name of the first author was erroneously printed “Ying Ling”. The correct spelling is “Ying Li”.

Ying Li, Xingjin Mao

Backmatter

Weitere Informationen

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.

Whitepaper

- ANZEIGE -

Globales Erdungssystem in urbanen Kabelnetzen

Bedingt durch die Altersstruktur vieler Kabelverteilnetze mit der damit verbundenen verminderten Isolationsfestigkeit oder durch fortschreitenden Kabelausbau ist es immer häufiger erforderlich, anstelle der Resonanz-Sternpunktserdung alternative Konzepte für die Sternpunktsbehandlung umzusetzen. Die damit verbundenen Fehlerortungskonzepte bzw. die Erhöhung der Restströme im Erdschlussfall führen jedoch aufgrund der hohen Fehlerströme zu neuen Anforderungen an die Erdungs- und Fehlerstromrückleitungs-Systeme. Lesen Sie hier über die Auswirkung von leitfähigen Strukturen auf die Stromaufteilung sowie die Potentialverhältnisse in urbanen Kabelnetzen bei stromstarken Erdschlüssen. Jetzt gratis downloaden!

Bildnachweise