模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2019 Vol.32 Issue.5, Published 2019-05-25
Papers and Reports    Researches and Applications    Surveys and Reviews   
   
Papers and Reports
385 Information Exchange Particle Swarm Optimization for Multitasking
CHENG Meiying1, QIAN Qian2, NI Zhiwei3, ZHU Xuhui3
Different from the existing cloud platform and parallel computer, the implicit parallelism of particle swarm optimization(PSO) is fully exploited in this paper. Two information exchange strategies, within-task information transfer and between-task information transfer, are involved. Moreover, factorial rank, scalar fitness and skill factor are introduced into PSO for multitasking. In each iteration, the most appropriate individuals are used to solve the most suitable task, and information exchange PSO for multitasking(IEPSOM) is proposed. Multitasking function optimization problems, multitasking multiple constraints Engineering design cases and multitasking key evaluation system constructing problems are involved to verify the performance of IEPSOM. Experimental results reveal that in IEPSOM multitasking environment, information transfer enhances the solutions quality and speeds up the convergence.
2019 Vol. 32 (5): 385-397 [Abstract] ( 584 ) [HTML 0KB] [ PDF 1140KB] ( 483 )
398 MS and PAN Image Fusion Algorithm Based on PST Phase Constraint and Sparse Representation
WANG Xianghai1,2, BAI Shifu1, LI Zhi1, SONG Ruoxi2, TAO Jingzhe2
In the remote sensing image fusion based on multi-spectral(MS) image and panchromatic(PAN) image, effective extracting the texture feature information of PAN and injecting targeted information into MS image are crucial to the high quality of image fusion. Therefore, the MS and PAN image pansharpening algorithm based on phase constraint of phase stretch transform(PST) and sparse representation is proposed in this paper. Firstly, the MS and PAN images are filtered by Gaussian filter. For the low and medium frequency information, the fusion weight constraint is obtained by the phase difference of high frequency based on the sensitivity of the PST phase difference to the edge and texture region in the image. For the high frequency information, a training dictionary is obtained by learning the high frequency information of the PAN image, and the dictionary is used to sparsely represent and fuse the high frequency information of MS and PAN images, therefore the accuracy of high frequency fusion is improved. The proposed algorithm overcomes the poor fusion effect of traditional fusion methods on the edge texture region and the distortion of spectral information, achieves better fusion result. A large number of simulation experiments verify the effectiveness of the proposed method.
2019 Vol. 32 (5): 398-408 [Abstract] ( 458 ) [HTML 0KB] [ PDF 3336KB] ( 345 )
409 Local Gaussian Distribution Fitting Energy Model with Fractional Differential
CHU Jun1 , YU Jiajia2 , MIAO Jun1, ZHANG Guimei1
Local Gaussian distribution fitting energy(LGDF) model lacks global information and is sensitive to the selection of initial contour curve. Especially in the segmentation of images with weak edges and textures, it is easy to fall into the local extremum and is poorly robust to noise. To solve the problems, an LGDF model with fractional differential is proposed. A global Grümwald-Letnikov fractional gradient fitting term is introduced into LGDF model to enhance the gradient information of the weak edge and texture regions and improve the robustness to initial contour curve and noise. The coefficients of global and local terms are determined by adaptive weighting function to improve the efficiency and accuracy of gray-scale inhomogeneous image segmentation. The adaptive fractional order function is constructed according to gradient modulus, information entropy and contrast of the image to improve the segmentation efficiency. Both theoretical analysis and experiments show that the model can be used for the segmentation of the gray-scale inhomogeneous images and the images with weak texture and weak edge. Experiments on synthetic and real images show that the proposed model improves the accuracy and efficiency of image segmentation.
2019 Vol. 32 (5): 409-419 [Abstract] ( 523 ) [HTML 0KB] [ PDF 1993KB] ( 285 )
420 Spectral Clustering Algorithm Based on Weighted Ensemble Nyström Sampling
QIU Yunfei1, LIU Chang1

Since most Nyström methods have problems of unstable clustering effect and weak representativeness in spectral clustering application,a spectral clustering algorithm based on weighted ensemble Nyström sampling is proposed. Firstly, the statistical leverage score is used to distinguish the importance of data and the data are weighted. Then, based on these weights, the weighted K-means center point sampling is used to obtain multiple sets of sampling points. The integration framework is introduced, and the approximate kernel matrix is constructed using the cluster parallel operation Nyström method. Finally, the approximate kernel is determined by the ridge regression method. The matrices are combined to produce a more accurate low rank approximation than that by standard Nyström method. Experiments on UCI datasets demonstrate that the proposed algorithm achieves better clustering results.
2019 Vol. 32 (5): 420-428 [Abstract] ( 465 ) [HTML 1KB] [ PDF 682KB] ( 330 )
429 Spatial-Temporal Fusion Algorithm for Remote Sensing Images Based on Multi-input Dense Connected Neural Network
YAO Kaixuan1, CAO Feilong1
To solve the spatial-temporal fusion problem of images of surface reflectivity remote sensing satellites Landsat and MODIS, a spatial-temporal fusion algorithm for remote sensing images based on multi-input dense connected neural network is proposed. Firstly, a multi-input dense connected neural network is put forward to study the remote sensing images containing the difference information between continuous moments. Then, two transition images learned from the network are fused with the two known high spatial resolution images based on the difference similarity hypothesis to obtain the final predicted images. According to the fusion experiment of Landsat remote sensing images and MODIS remote sensing images, the proposed algorithm produces promising results in each quantitative index. The final predicted image by the proposed algorithm is more robust to noise with better recovered detail information.
2019 Vol. 32 (5): 429-435 [Abstract] ( 504 ) [HTML 0KB] [ PDF 1173KB] ( 341 )
Surveys and Reviews
436 Progress of Palmprint Recognition: A Review
ZHONG Dexing1, ZHU Jinsong1, DU Xuefeng1
Palmprint images contain rich features and they can be easily combined with hand dorsal veins, finger-knuckle-prints and hand shapes to form multimodal features. Thus, palmprint recognition becomes a hot topic in the field of biometric recognition. In this paper, the basic process of palmprint recognition is discussed from three aspects: the collection of palmprint images, region of interest detection, and feature extraction and matching. Multimodal methods based on fusion of different features are also explored. Besides, on the basis of feature extraction means, palmprint recognition algorithms are roughly divided into hand-craft based algorithms including encoding, structure and statistics, and subspace and feature learning based algorithms including machine learning and deep learning. The algorithms are compared and analyzed in detail. Finally, challenges and future perspectives in palmprint recognition are discussed, especially palmprint recognition system in complex environment across different devices.
2019 Vol. 32 (5): 436-445 [Abstract] ( 1405 ) [HTML 0KB] [ PDF 835KB] ( 1028 )
Researches and Applications
446 Wasserstein Distance Based Hierarchical Attention Model for Cross-Domain Sentiment Classification
DU Yongping1, HE Meng1, ZHAO Xiaozheng1
The task of cross-domain sentiment classification is to analyze the sentiment orientation of the target domain lacking labeled data using the source-domain data with sentiment labels. A hierarchical attention model based on Wasserstein distance is proposed in this paper. The hierarchical model is used for feature extraction by combining attention mechanism, and Wasserstein distance is used as the domain difference metric to automatically capture the domain-sharing features through adversarial training. Further auxiliary task is constructed to capture the domain-special features cooccurring with domain-sharing features. These two kinds of features are united to complete the cross-domain sentiment classification task. The experimental results on Amazon datasets demonstrate that the proposed model achieves a higher accuracy and a better stability on different cross-domain pairs.
2019 Vol. 32 (5): 446-454 [Abstract] ( 520 ) [HTML 1KB] [ PDF 791KB] ( 539 )
455 Dynamic Multiple-level Semantic Extraction Model Based on External Knowledge
JIANG Wenchao1,2, ZHUANG Zhigang1, TU Xuping2, LI Chuanjie3, LIU Haibo1,
To resolve the problems of semantic understanding and answer extraction in complex multiple context machine reading comprehension environments, a dynamic multiple-level semantic extraction model based on external knowledge is presented. Firstly, the optimized gated single cyclic neural network model is utilized to match the text as well as the problem set. Then, the dynamic multiple-dimension bidirectional attention mechanism analysis is implemented on the text and the problem set respectively to improve the semantic matching precision. Next, a dynamic pointer network is utilized to determine the rank of the answers to the questions. Finally, the candidate answers are sorted based on external knowledge and experiences, and the precision of the final answer is improved further. The experimental results show that problem-answer matching accuracy of the proposed model is significantly improved compared with the mainstream models. Furthermore, the proposed model obtains higher robustness in complex reading comprehension tasks in different application scenes.
2019 Vol. 32 (5): 455-462 [Abstract] ( 364 ) [HTML 0KB] [ PDF 646KB] ( 794 )
463 Travel Attractions Recommendation Based on Trajectory Mining Representation Model
ZHANG Shunyao1, CHANG Liang2, GU Tianlong1, BIN Chenzhong2,
SUN Yanpeng3, ZHU Guiming1, JIA Zhonghao1
A recommendation method based on the gated recurrent unit trajectory mining representation model(GRU-TMRM) is proposed to solve the problems of data sparsity and cold start in content based and collaborate filter based recommendation method, as well as the problem of ignoring rich semantics of travel track in track mining method. To take full advantage of semantics information contained in travel track, GRU-TMRM is designed. With GRU-TMRM, historical tracks of visitors can be modeled for providing personalized attractions recommendation. Experiments on real travel track dataset show that the proposed method effectively improves the accuracy and quality of recommendation compared with the widely used baseline method.
2019 Vol. 32 (5): 463-471 [Abstract] ( 533 ) [HTML 0KB] [ PDF 1146KB] ( 465 )
472 Semantic Segmentation Method for Complex Traffic Scene Based on DenseNet
JIANG Bin1, TU Wenxuan1, YANG Chao1, LIU Hongyu1, ZHAO Zilong1
An end-to-end multi-scale semantic segmentation model based on fully convolutional DenseNet is proposed, aiming at the problems of traditional semantic segmentation methods for street scene, such as the large number of parameters and low computational efficiency and precision. Firstly, convolution layers embedded with hybrid dilation convolution are stacked to establish a dense module, and then the modules are cascaded along channel dimension to extract features. Next, multi-scale visual information regarded as supervised signals are transferred back to original channels. Finally, the prediction results are obtained by bilinear interpolation method. Experimental results on Cityscapes dataset demonstrate that the proposed method achieves an efficient segmentation and performs a better accuracy for street scene parsing.
2019 Vol. 32 (5): 472-480 [Abstract] ( 732 ) [HTML 0KB] [ PDF 2007KB] ( 691 )
模式识别与人工智能
 

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