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....

2017 Vol.30 Issue.6, Published 2017-06-30 Papers and Reports    Researches and Applications        Papers and Reports 481 Laplacian Sparse Coding by Incorporating Locality and Non-negativity for Image Classification WAN Yuan, SHI Ying, WU Kefeng, CHEN Xiaoli The local relationship between features of images is not taken into account in the traditional sparse coding, it can lead to the instability of encoding. Moreover, some effective features may not be retained via the subtraction operation in the optimization procedure. Aiming at these two problems, a method is proposed, named Laplacian sparse coding by incorporating locality and non-negativity(LN-LSC)for image classification. Firstly, bases near to the local features are chosen to constrain the codes. Then, non-negativity is introduced in Laplacian sparse coding by non-negative matrix factorization. Finally, spatial pyramid division and max pooling are utilized to generate the final representation of images in the pooling step. Multi-class linear SVM is adopted for image classification. The local information between features is preserved by the proposed method, and the offsetting between features is also avoided. Thus, more features are utilized for coding, and the instability of the coding is overcome. Experiments on four public image datasets show the classification accuracy of LN-LSC is higher than that of the state-of-the-art sparse coding algorithms. 2017 Vol. 30 (6): 481-488 [Abstract] ( 609 ) [HTML 1KB] [ PDF 1060KB] ( 525 ) 489 Algorithm for Maximal Successful Coalition Generation with Goals Preferences ZHANG Guofu, DU Xiaodong, SU Zhaopin, JIANG Jianguo In the traditional research on coalitional resource games(CRGs), it is assumed that an agent can respond to any goal, even if the agent is not interested in the goal at all. In this paper, a natural variation of CRGs with goals preferences is discussed. An agent only contributes its resources to the goals in its own goal (or interest) set. For this purpose, an improved CRGs model is firstly proposed on the basis of goals preferences. Moreover, a two-dimensional binary encoding based algorithm for maximal successful coalition (MAXSC) generation is designed and a heuristic algorithm is developed to resolve the potential conflicts of agents scrambling for scarce resources. Finally, the proposed approach is compared with the previous algorithms for the MAXSC problem.The results demonstrate the effectiveness of the proposed approach. 2017 Vol. 30 (6): 489-498 [Abstract] ( 430 ) [HTML 1KB] [ PDF 939KB] ( 732 ) 499 Multi-scale Collaborative Coupled Metric Learning Method Based on the Fusion of Class and Structure Information ZOU Guofeng, FU Guixia, GAO Mingliang, YIN Liju, WANG Kejun Aiming at the elements matching problem in different scale space sets, multi-scale collaborative coupled metric learning method based on the fusion of class and structure information is proposed. Firstly, the correlation matrix is constructed under the guidance of class information and structure information of sample distribution. The class information is significant for supervision and the structure information is the auxiliary supervision information. The linear and nonlinear optimization objective equations are constructed based on the correlation matrix. By solving the optimization objective equation, the samples are transformed from different scale space datasets into a unified public space for distance measurement. The experimental results of face recognition show that the nonlinear collaborative coupled metric is an effective measurement method and it is simple and convenient with a higher recognition rate. 2017 Vol. 30 (6): 499-508 [Abstract] ( 531 ) [HTML 1KB] [ PDF 976KB] ( 464 ) 509 Preference-Inspired Co-evolutionary Algorithm Based on Hybrid Domination Strategy WANG Liping, DU Jiejie, QIU Feiyue, JIANG Bo The preference-inspired co-evolutionary algorithm employing goal vectors can not identify the Pareto dominance relationship of candidate solutions at the same fitness level, and the obtained solutions are unevenly distributed along the Pareto front. Aiming at these problems, preference-inspired co-evolutionary algorithm based on hybrid domination strategy(E-PICEA-g) is proposed in this paper. Firstly, Pareto dominance sorting on population is conducted, and then the candidate solutions fitness values are calculated to reduce the proportion of non-dominated solutions in the population and increase the selection pressure. Meanwhile, the distance between candidate solutions and ideal point is considered to punish the candidate solutions at the same fitness level but far from the ideal point. Thus, the obtained solutions are made to distribute evenly along the Pareto optimal front. Experimental results on 12 multi-objective optimization functions demonstrate that the proposed algorithm acquires solutions with high quality on most of the test functions. 2017 Vol. 30 (6): 509-519 [Abstract] ( 543 ) [HTML 1KB] [ PDF 1289KB] ( 477 ) 520 Indoor Scene Classification Algorithm Based on Information Enhancement of Vision Sensitive Area SHI Jing, ZHU Hong, WANG Jing, XUE Shan In the indoor scene classification, the classification accuracy is affected by various interference factors caused by the complexity and diversity of the scene structure itself. Aiming at these problems, an indoor scene classification algorithm based on the information enhancement of visual sensitive area is proposed in this paper. By fusing the local features and the global features based on the visual sensitive region information, the multi-scale space-frequency fusion feature is constructed to classify the indoor scenes correctly. Experimental results on 3 testing sets show that the proposed algorithm obtains good classification results on different scene classification datasets with strong applicability. 2017 Vol. 30 (6): 520-529 [Abstract] ( 449 ) [HTML 1KB] [ PDF 1792KB] ( 508 ) Researches and Applications 530 Neighborhood Repulsed Metric Learning for Kinship Verification Based on Local Feature Fusion HU Zhengping, GUO Zengjie, WANG Meng, SUN Zhe To solve the problem of kinship verification of facial image, an algorithm for neighborhood repulsed metric learning based on local feature fusion is proposed. Firstly, texture and skin color features are extracted from the key areas of the face images. Then, the feature fusion method is proposed. Finally, the metric learning method is introduced to learn a transformational matrix capable of making the distance between the samples with kinship smaller and the distance between the samples of non-kin larger. The prior knowledge of the similarity degree of existing data samples is utilized to learn the best similarity measure to describe the similarity of kinship samples better. The experimental results on KinFaceW-I and KinFaceW-II databases demonstrate the efficiency of the proposed algorithm. 2017 Vol. 30 (6): 530-537 [Abstract] ( 477 ) [HTML 1KB] [ PDF 948KB] ( 423 ) 538 Dynamic Search Firefly Algorithm Based on Improved Attraction LI Rongyu, CHEN Qingqian, CHEN Siyuan Due to the decreasing attraction in solving high-dimensional optimization problems, the basic firefly algorithm easily falls into local optimum with low accuracy. Aiming at this problem, the dynamic search firefly algorithm based on improved attraction(ADFA) is proposed in this paper. The minimum attraction concept is presented to enhance the individual communication. By adding the optimal value of objective function, the step size can be adjusted adaptively. Finally, ten benchmark functions from CEC2014 are exploited to validate ADFA. The simulation results show that ADFA obtains higher convergence rate and better accuracy. 2017 Vol. 30 (6): 538-548 [Abstract] ( 461 ) [HTML 1KB] [ PDF 834KB] ( 339 ) 549 Classification Method of fMRI Data Based on Convolutional Neural Network ZHANG Zhaochen, JI Junzhong Since classification method of functional magnetic resonance imaging(fMRI) data can not effectively extract the local features, the classification accuracy is seriously affected. To solve the problem, a classification model of fMRI data based on convolutional neural network(CNN) is presented. Firstly, a CNN structure is designed, and a restricted boltzmann machine(RBM) model is constructed by means of the convolution kernel size. Then, the interested region voxels in fMRI data are employed to construct and form input data to pre-train RBM, and the relative transformation of the obtained weight matrix is executed to initialize CNN parameters. Finally, the final classification model is obtained by training the whole initialized model. The results on Haxby and LPD datasets show that the proposed model effectively improves the classification accuracy of fMRI data. 2017 Vol. 30 (6): 549-558 [Abstract] ( 1022 ) [HTML 1KB] [ PDF 811KB] ( 755 ) 559 Multilingual Documents Clustering Algorithm Based on Parallel Information Bottleneck YAN Xiaoqiang, LU Yaoen, LOU Zhengzheng, YE Yangdong The potential complementation between different languages is ignored while traditional clustering algorithms discover the hidden structures in document collection. Thus, the latent information in the collection can not be reflected by the obtained patterns. Aiming at this problem, multilingual document clustering algorithm based on parallel information bottleneck(ML-IB) is proposed. Firstly, the relevant variables of multiple language information are constructed according to the bag-of-words model. Then,the multiple relevant variables are incorporated into the parallel information bottleneck, and the relevant information between data patterns and multiple relevant variables is preserved maximally. Finally, to optimize the objective function of ML-IB, a draw and merge method based on information theory is proposed to guarantee the convergence of ML-IB to a local optimal solution. Extensive experimental results on multilingual document datasets show that the proposed algorithm significantly outperform the state-of-the-art single and multilingual clustering methods. 2017 Vol. 30 (6): 559-568 [Abstract] ( 503 ) [HTML 1KB] [ PDF 640KB] ( 783 ) 569 Sub-regional and Multi-classifier Vehicle Detection Based on Haar-like and MB-LBP Features ZHU Bin, WANG Shaoping, LIANG Huawei, YUAN Sheng, YANG Jing, HUANG Junjie The pre-collision system in the advanced vehicle-assisted driving system serves to detect both the forward vehicle to prevent the rear-end collision and the adjacent lane oblique lateral vehicle to realize the forecast of its potential lane change and confluence behavior, providing real-time warning function. In this paper, a method of vehicle detection with multiple characteristics of sub-region fusion is proposed to solve the obvious differences between forward vehicle and oblique lateral vehicle. The proposed method is utilized to detect the vehicles at different distances, respectively, to conduct image downsampling with different degrees. Consequently, the real-time performance of the detection system is improved. The field test of various traffic scenes indicates that the proposed method can detect the forward vehicle and the oblique side vehicle stably and accurately in real time. In good driving environment, the proposed method can achieve high recall rate and accuracy rate. 2017 Vol. 30 (6): 569-576 [Abstract] ( 622 ) [HTML 1KB] [ PDF 1390KB] ( 621 )

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