粒子流粒子滤波检测前跟踪方法
|
摘要
检测前跟踪通过积累多帧量测以检测和跟踪微弱目标。积累的关键在于对目标后验密度的准确表示。传统粒子滤波器过于依赖建议密度,因而对目标后验密度的表示不够准确。新提出的粒子流滤波器能够准确表示目标后验密度,但无法实现量测的帧间积累。为此,本文提出一种在粒子滤波框架下结合粒子流的检测前跟踪方法:采用粒子滤波框架实现多帧量测积累,并在每一帧内采用Localized Exact Daum-Huang粒子流表示目标后验密度,以提升量测积累效果。我们通过Rayleigh杂波下Swerling1型起伏目标的检测和跟踪实验证明了所提算法的性能。
The track-before-detect strategy detects and tracks weak targets by the integration of measurements in multiple frames. The key step to the integration is the accurate characterization of the target posterior. The traditional particle filter relies too much on the proposal density, and thus the characterization is not exact enough. The newly presented particle flow filter can represent the target posterior exactly. However, it neglects the multi-frame integration. Therefore, in this paper, a novel track-before-detect scheme is proposed, which incorporates a particle flow filter into an encompassing particle filter framework. The particle filter is used for multi-frame measurement integration, and within each frame the Localized Exact Daum-Huang filter is used to represent the target posterior to improve the effect of integration. The performance of the proposed algorithm is evaluated by simulations of a Swerling 1 fluctuating target detecting and tracking in Rayleigh clutter.
The track-before-detect strategy detects and tracks weak targets by the integration of measurements in multiple frames. The key step to the integration is the accurate characterization of the target posterior. The traditional particle filter relies too much on the proposal density, and thus the characterization is not exact enough. The newly presented particle flow filter can represent the target posterior exactly. However, it neglects the multi-frame integration. Therefore, in this paper, a novel track-before-detect scheme is proposed, which incorporates a particle flow filter into an encompassing particle filter framework. The particle filter is used for multi-frame measurement integration, and within each frame the Localized Exact Daum-Huang filter is used to represent the target posterior to improve the effect of integration. The performance of the proposed algorithm is evaluated by simulations of a Swerling 1 fluctuating target detecting and tracking in Rayleigh clutter.
引文
[1] Davey S J,Rutten M G,Gordon A N J.Track-before-detect techniques:integrated tracking,classification,and sensor management:theory and applications[M].USA,John Wiley & Sons,2013:311-318.
[2] Berry P,Venkataraman K,Rosenberg L.Adaptive detection of low-observable targets in correlated sea clutter using Bayesian track-before-detect [C]//in Proc.IEEE Radar Conf.,Seattle,WA,USA,2017:398- 403.
[3] Northardt T,Nardone S C.Track-before-detect bearings-only localization performance in complex passive sonar scenarios:a case study[J].IEEE Journ.Oceanic Engineer.(preprint).
[4] Davey S J,Rutten M G,Cheung B.A comparison of detection performance for several Track-before-detect algorithms[J].Eura.Journ.Advan.Signal Process.,2007,2008(1):1-10.
[5] Jiang H,Yi W,Kirubarajan T,et al.Multi-frame radar detection of fluctuating targets using phase information[J].IEEE Trans Aerosp.Electr.Syst,2017,53(2):736-749.
[6] Sahin G,Demirekler M.A Multi-dimensional hough transform algorithm based on unscented transform as a track-before-detect method[C]//in Proc.Int.Conf.Inf.Fusion.Salamanca,Spain,2014:1- 8.
[7] Ciuonzo D,Willett P K,Barshalom Y.Tracking the tracker from its passive sonar ML-PDA estimates [J].IEEE Trans.Aerosp.Electron.Syst.,2014,50(1):573-590.
[8] Boers Y,Driessen J N.Multitarget particle filter track before detect application[J].IEE Proc.Radar Sonar & Navig.,2004,151(6):351-357.
[9] Malcolm R,David S.Particle filter for track-before-detect of a target with unknown amplitude viewed against a structured scene[J].IET Radar,Sonar & Navig.,2018,12(6):603- 609.
[10] 柳超,李秀友,黄勇.优化的多模型粒子滤波机动微弱目标检测前跟踪方法[J].信号处理,2015,31(9):1131-1137.Liu Chao,Li Xiuyou,Huang Yong.Optimized multiple model particle filter track-before-detect algorithm for maneuvering weak target[J].Journal of Signal Processing,2015,31(9):1131-1137.(in Chinese)
[11] 王娜,谭顺成,王国宏.目标数未知时基于粒子滤波的多目标TBD方法[J].信号处理,2017,33(9):1248-1257.Wang Na,Tan Shuncheng,Wang Guohong.Multitarget particle filter track-before-detect algorithm with unknown target number[J].Journal of Signal Processing,2017,33(9):1248-1257.(in Chinese)
[12] Daum F,Huang J.Particle degeneracy:root cause and solution [C]//Proc.SPIE Conf.Signal Process.,Sensor Fusion and Target Recog.,Orlando,FL,2011,80500W.
[13] Li Y,Coates M.Particle filtering with invertible particle flow [J].IEEE Trans.Signal Process.,2017,65(15):4102- 4116.
[14] Doucet A,De Freitas N,Murphy K,et al.Rao-Blackwellised particle filtering for dynamic Bayesian networks[C]//Proc.Conf.Uncertainty Artif.Intell.,San Francisco,CA,USA,2000:176-183.
[15] Doucet A,Freitas N D,Wan E.The unscented particle filter[C]//in Proc.Neural Inf.Process.Syst.,Denver,CO,USA,Dec.2000:584-590.
[16] Pitt M,Shephard N.Filtering via simulation:auxiliary particle filters[J].Journ.Amer.Statist.Assoc.,1999,94(446):590-599.
[17] Septier F,Peters G W.Langevin and Hamiltonian based sequential MCMC for efficient Bayesian filtering in high-dimensional spaces[J].IEEE Journ.Selec.Topics Signal Process.,2016,10(2):312-327.
[18] Hanebeck U D,Briechle K,Rauh A.Progressive Bayes:A new framework for nonlinear state estimation [C]//in Proc.SPIE,2003(5099):256-267.
[19] Daum F,Huang J.Nonlinear filters with log-homotopy[C]//in Proc.SPIE,vol.6699,Sep.2007,Art.no.669918.
[20] Daum F,Huang J,Noushin A.Exact particle flow for nonlinear filters [J]//in Proc.SPIE,vol.7697,Apr.2010:769704.
[21] Ding T,Coates M.Implementation of the Daum-Huang exact flow particle filter[C]//in Proc.IEEE Statist.Signal Process.Workshop,Ann Arbor,MI,USA,Aug.2012:257-260.
[22] Reich S.A dynamical systems framework for intermittent data assimilation[J].Bit Numerical Mathematics,2011,51(1):235-249.
[23] Reich S.A Gaussian-mixture ensemble transform filter[J].Quarter.Journ.Royal Meteoro.Society,2012,138(662):222-233.
[24] Kreucher C,Bell K.A Geodesic flow particle filter for non-thresholded measurements[C]//in Proc.IEEE Radar Conf.,Seattle,WA,USA,2017:891- 896.
[25] Bunch P,Godsill S.Approximations of the optimal importance density using Gaussian particle flow importance sampling[J].Journ.Ame.Statist.Assoc.,2014,111(514):748-762.
[26] Daum F,Huang J.Particle flow for nonlinear filters,Bayesian decisions and transport[C]//in Proc.IEEE Int.Conf.Inf.Fusion,Istanbul,Turkey,2013:1072-1079.
[27] Punithakumar K,Kirubarajan T,Sinha,A.A sequential Monte Carlo probability hypothesis density algorithm for multitarget track-before-detect[J].Proc.SPIE-Int.Society Optical Engineer.,2005:59131S-59131S- 8.
[2] Berry P,Venkataraman K,Rosenberg L.Adaptive detection of low-observable targets in correlated sea clutter using Bayesian track-before-detect [C]//in Proc.IEEE Radar Conf.,Seattle,WA,USA,2017:398- 403.
[3] Northardt T,Nardone S C.Track-before-detect bearings-only localization performance in complex passive sonar scenarios:a case study[J].IEEE Journ.Oceanic Engineer.(preprint).
[4] Davey S J,Rutten M G,Cheung B.A comparison of detection performance for several Track-before-detect algorithms[J].Eura.Journ.Advan.Signal Process.,2007,2008(1):1-10.
[5] Jiang H,Yi W,Kirubarajan T,et al.Multi-frame radar detection of fluctuating targets using phase information[J].IEEE Trans Aerosp.Electr.Syst,2017,53(2):736-749.
[6] Sahin G,Demirekler M.A Multi-dimensional hough transform algorithm based on unscented transform as a track-before-detect method[C]//in Proc.Int.Conf.Inf.Fusion.Salamanca,Spain,2014:1- 8.
[7] Ciuonzo D,Willett P K,Barshalom Y.Tracking the tracker from its passive sonar ML-PDA estimates [J].IEEE Trans.Aerosp.Electron.Syst.,2014,50(1):573-590.
[8] Boers Y,Driessen J N.Multitarget particle filter track before detect application[J].IEE Proc.Radar Sonar & Navig.,2004,151(6):351-357.
[9] Malcolm R,David S.Particle filter for track-before-detect of a target with unknown amplitude viewed against a structured scene[J].IET Radar,Sonar & Navig.,2018,12(6):603- 609.
[10] 柳超,李秀友,黄勇.优化的多模型粒子滤波机动微弱目标检测前跟踪方法[J].信号处理,2015,31(9):1131-1137.Liu Chao,Li Xiuyou,Huang Yong.Optimized multiple model particle filter track-before-detect algorithm for maneuvering weak target[J].Journal of Signal Processing,2015,31(9):1131-1137.(in Chinese)
[11] 王娜,谭顺成,王国宏.目标数未知时基于粒子滤波的多目标TBD方法[J].信号处理,2017,33(9):1248-1257.Wang Na,Tan Shuncheng,Wang Guohong.Multitarget particle filter track-before-detect algorithm with unknown target number[J].Journal of Signal Processing,2017,33(9):1248-1257.(in Chinese)
[12] Daum F,Huang J.Particle degeneracy:root cause and solution [C]//Proc.SPIE Conf.Signal Process.,Sensor Fusion and Target Recog.,Orlando,FL,2011,80500W.
[13] Li Y,Coates M.Particle filtering with invertible particle flow [J].IEEE Trans.Signal Process.,2017,65(15):4102- 4116.
[14] Doucet A,De Freitas N,Murphy K,et al.Rao-Blackwellised particle filtering for dynamic Bayesian networks[C]//Proc.Conf.Uncertainty Artif.Intell.,San Francisco,CA,USA,2000:176-183.
[15] Doucet A,Freitas N D,Wan E.The unscented particle filter[C]//in Proc.Neural Inf.Process.Syst.,Denver,CO,USA,Dec.2000:584-590.
[16] Pitt M,Shephard N.Filtering via simulation:auxiliary particle filters[J].Journ.Amer.Statist.Assoc.,1999,94(446):590-599.
[17] Septier F,Peters G W.Langevin and Hamiltonian based sequential MCMC for efficient Bayesian filtering in high-dimensional spaces[J].IEEE Journ.Selec.Topics Signal Process.,2016,10(2):312-327.
[18] Hanebeck U D,Briechle K,Rauh A.Progressive Bayes:A new framework for nonlinear state estimation [C]//in Proc.SPIE,2003(5099):256-267.
[19] Daum F,Huang J.Nonlinear filters with log-homotopy[C]//in Proc.SPIE,vol.6699,Sep.2007,Art.no.669918.
[20] Daum F,Huang J,Noushin A.Exact particle flow for nonlinear filters [J]//in Proc.SPIE,vol.7697,Apr.2010:769704.
[21] Ding T,Coates M.Implementation of the Daum-Huang exact flow particle filter[C]//in Proc.IEEE Statist.Signal Process.Workshop,Ann Arbor,MI,USA,Aug.2012:257-260.
[22] Reich S.A dynamical systems framework for intermittent data assimilation[J].Bit Numerical Mathematics,2011,51(1):235-249.
[23] Reich S.A Gaussian-mixture ensemble transform filter[J].Quarter.Journ.Royal Meteoro.Society,2012,138(662):222-233.
[24] Kreucher C,Bell K.A Geodesic flow particle filter for non-thresholded measurements[C]//in Proc.IEEE Radar Conf.,Seattle,WA,USA,2017:891- 896.
[25] Bunch P,Godsill S.Approximations of the optimal importance density using Gaussian particle flow importance sampling[J].Journ.Ame.Statist.Assoc.,2014,111(514):748-762.
[26] Daum F,Huang J.Particle flow for nonlinear filters,Bayesian decisions and transport[C]//in Proc.IEEE Int.Conf.Inf.Fusion,Istanbul,Turkey,2013:1072-1079.
[27] Punithakumar K,Kirubarajan T,Sinha,A.A sequential Monte Carlo probability hypothesis density algorithm for multitarget track-before-detect[J].Proc.SPIE-Int.Society Optical Engineer.,2005:59131S-59131S- 8.