基于二维特征矩阵的特征融合算法
|
摘要
为了提高无线传感器网络信息融合的效率,提出一种多传感器二维特征融合(2DFF)策略.将多个传感器标准化后的特征集组合成二维特征矩阵,引入图像压缩技术,包括二维主成分分析(2DPCA)及MatPCA对特征矩阵进行特征提取,实现特征融合.从理论上剖析该方法之所以能够有效地适用于特征融合,且区别于传统方法的内在本质.相比传统的特征融合方法,该方法能够获得更加精确的融合特征,提高信息融合的效率.基于实地采集的地面目标信号的实验结果表明,该方法既提高目标识别率,又降低了计算复杂度.
A strategy of two-dimensional feature fusion(2DFF) was proposed to improve efficiency of information fusion in wireless sensor networks(WSNs).The feature sets after normalized of multi-sensors were combined into a two-dimensional feature matrix.The techniques of image compression,two-dimensional principal component analysis(2DPCA) and MatPCA were generalized for feature extraction from two-dimensional feature matrix to achieve multi-sensors feature fusion.The inherent essence of this method used in feature fusion was analyzed further in theory.Compared to traditional feature fusion method,this method can obtain more accurate fused feature and improve efficiency of information fusion.The experiment results on the real signals of ground targets show that this method can increase classification accuracy and reduce computational complexity.
A strategy of two-dimensional feature fusion(2DFF) was proposed to improve efficiency of information fusion in wireless sensor networks(WSNs).The feature sets after normalized of multi-sensors were combined into a two-dimensional feature matrix.The techniques of image compression,two-dimensional principal component analysis(2DPCA) and MatPCA were generalized for feature extraction from two-dimensional feature matrix to achieve multi-sensors feature fusion.The inherent essence of this method used in feature fusion was analyzed further in theory.Compared to traditional feature fusion method,this method can obtain more accurate fused feature and improve efficiency of information fusion.The experiment results on the real signals of ground targets show that this method can increase classification accuracy and reduce computational complexity.
引文
[1]HALL D L,LLINAS J.Handbook of multisensor datafsuion[M].New York:CRC Press,2001:33-66.
[2]MALHOTRA B,NIKOLAIDIS I,HARMS J.Distrib-uted classification of acoustic taargets in wireless audio-sensor networks[J].Computer Networks,2008,52(13):2582-2593.
[3]BARDWAJ A A,ANANDARAJ M,KAPIL K,et al.Multi sensor data fusion methods using sensor data com-pression and estimated weights[C]∥International Con-ference on Signal Processing,Communications and Net-working,2008.ICSCN’08.Chennai India:Madras Insti-tute of Technology,Anna University,2008,(4/6):250-254.
[4]KUNCHEVA L I,BEZDEK J C,DUIN R P W.Deci-sion templates for multiple classifier fusion:an experi-mental comparison[J].Pattern Recognition,2001,34(2):299-314.
[5]PAN Qiang,WEI Jian-ming,CAO Hong-bin,et al.Improved DS acoustic-seismic modality fusion forground-moving target classification in wireless sensornetworks[J].Pattern Recognition Letters,2007,28(16):2419-2426.
[6]曹红兵,魏建明,刘海涛.无线传感网中多传感器特征融合算法研究[J].电子与信息学报,2010,32(1):166-171.CAO Hong-bin,WEI Jian-ming,LIU Hai-tao.Re-search on multi-Sensor feature fusion algorithms in wire-less sensor networks[J].Journal of Electronics&In-formation Technology,2010,32(1):166-171.
[7]ARNAZ M,ROBERT X G.PCA-based feature selec-tion scheme for machine defect classification[J].IEEETransactions on Instrumentation and Measurement,2004,53(6):1517-1525.
[8]TURBAN-SAYAN G.Real time electromagnetic targetclassification using a novel feature extraction techniquewith PCA-based fusion[J].IEEE Transactions on An-tennas and Propagation,2005,53(2):766-776.
[9]MICHAEL B R,INDRA A.Classification and regional-ization through kernel principal component analysis[J].Physics and Chemistry of the Earth,2010(35):316-328.
[10]SUN Bing-yu,ZHANG Xiao-ming,LI Jiu-yong,et al.Feature fusion using locally linear embedding for classi-fication[J].IEEE Transactions on Neural Networks,2010,21(1):163-168.
[11]YANG Jian,YANG Jing-yu.Generalized K–Ltransform based combined feature extraction[J].Pat-tern Recognition,2002,35(1):295-297.
[12]YANG Jian,YANG Jing-yu,ZHANG David,et al.Feature fusion:parallel strategy vs.serial strategy[J].Pattern Recognition,2003,36(6):1369-1381.
[13]SUN Quan-sen,ZENG Sheng-gen,LIU Yan,et al.Anew method of feature fusion and its application in im-age recognition[J].Pattern Recognition,2005,38(12):2437-2448.
[14]YANG Jian,ZHANG David,Alejandro F F,et al.Two-dimensional PCA:a new approach to appearance-based face representation and recognition[J].IEEETransactions on Pattern Analysis and Machine Intelli-gence,2004,26(1):131-137.
[15]CHEN Song-can,ZHU Yu-lian,ZHANG Dao-qiang,et al.Feature extraction approaches based on matrixpattern:MatPCA and MatFLDA[J].Pattern Recogni-tion Letters,2005,26(8):1157-1167.
[16]YANG Jian,YANG Jing-yu.From image vector tomatrix:a straightforward image projection technique-IMPCA vs.PCA[J].Pattern Recognition,2002,35(9):1997-1999.
[17]李舜酩,李香莲.振动信号的现代分析技术与应用[M].1版.北京:国防工业出版社,2008:57-117.
[18]聂为荣,朱继南.运动目标地震动信号的时频特征分析[J].南京理工大学学报,2002,26(5):478-481.NIE Wei-rong,ZHU Ji-nan.Eigenvector Analysis inTime and frequency of seismic signals of moving target[J].Journal of Nanjing University of Science and Tech-nology,2002,26(5):478-481.
[2]MALHOTRA B,NIKOLAIDIS I,HARMS J.Distrib-uted classification of acoustic taargets in wireless audio-sensor networks[J].Computer Networks,2008,52(13):2582-2593.
[3]BARDWAJ A A,ANANDARAJ M,KAPIL K,et al.Multi sensor data fusion methods using sensor data com-pression and estimated weights[C]∥International Con-ference on Signal Processing,Communications and Net-working,2008.ICSCN’08.Chennai India:Madras Insti-tute of Technology,Anna University,2008,(4/6):250-254.
[4]KUNCHEVA L I,BEZDEK J C,DUIN R P W.Deci-sion templates for multiple classifier fusion:an experi-mental comparison[J].Pattern Recognition,2001,34(2):299-314.
[5]PAN Qiang,WEI Jian-ming,CAO Hong-bin,et al.Improved DS acoustic-seismic modality fusion forground-moving target classification in wireless sensornetworks[J].Pattern Recognition Letters,2007,28(16):2419-2426.
[6]曹红兵,魏建明,刘海涛.无线传感网中多传感器特征融合算法研究[J].电子与信息学报,2010,32(1):166-171.CAO Hong-bin,WEI Jian-ming,LIU Hai-tao.Re-search on multi-Sensor feature fusion algorithms in wire-less sensor networks[J].Journal of Electronics&In-formation Technology,2010,32(1):166-171.
[7]ARNAZ M,ROBERT X G.PCA-based feature selec-tion scheme for machine defect classification[J].IEEETransactions on Instrumentation and Measurement,2004,53(6):1517-1525.
[8]TURBAN-SAYAN G.Real time electromagnetic targetclassification using a novel feature extraction techniquewith PCA-based fusion[J].IEEE Transactions on An-tennas and Propagation,2005,53(2):766-776.
[9]MICHAEL B R,INDRA A.Classification and regional-ization through kernel principal component analysis[J].Physics and Chemistry of the Earth,2010(35):316-328.
[10]SUN Bing-yu,ZHANG Xiao-ming,LI Jiu-yong,et al.Feature fusion using locally linear embedding for classi-fication[J].IEEE Transactions on Neural Networks,2010,21(1):163-168.
[11]YANG Jian,YANG Jing-yu.Generalized K–Ltransform based combined feature extraction[J].Pat-tern Recognition,2002,35(1):295-297.
[12]YANG Jian,YANG Jing-yu,ZHANG David,et al.Feature fusion:parallel strategy vs.serial strategy[J].Pattern Recognition,2003,36(6):1369-1381.
[13]SUN Quan-sen,ZENG Sheng-gen,LIU Yan,et al.Anew method of feature fusion and its application in im-age recognition[J].Pattern Recognition,2005,38(12):2437-2448.
[14]YANG Jian,ZHANG David,Alejandro F F,et al.Two-dimensional PCA:a new approach to appearance-based face representation and recognition[J].IEEETransactions on Pattern Analysis and Machine Intelli-gence,2004,26(1):131-137.
[15]CHEN Song-can,ZHU Yu-lian,ZHANG Dao-qiang,et al.Feature extraction approaches based on matrixpattern:MatPCA and MatFLDA[J].Pattern Recogni-tion Letters,2005,26(8):1157-1167.
[16]YANG Jian,YANG Jing-yu.From image vector tomatrix:a straightforward image projection technique-IMPCA vs.PCA[J].Pattern Recognition,2002,35(9):1997-1999.
[17]李舜酩,李香莲.振动信号的现代分析技术与应用[M].1版.北京:国防工业出版社,2008:57-117.
[18]聂为荣,朱继南.运动目标地震动信号的时频特征分析[J].南京理工大学学报,2002,26(5):478-481.NIE Wei-rong,ZHU Ji-nan.Eigenvector Analysis inTime and frequency of seismic signals of moving target[J].Journal of Nanjing University of Science and Tech-nology,2002,26(5):478-481.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心