THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. 2016 Mar.; 27(3), 317325. http://dx.doi.org/10.5515/kjkiees.2016.27.3.317 ISSN 1226-3133 (Print)ISSN 2288-226X (Online) TBD Radar Tracking Using Particle Filter for Track-Before-Detect(TBD) 권지훈 강성철 곽노준 Ji-Hoon KwonSeung-Chul KangNo-Jun Kwak* 요약 --(TBD: Track Before Detect)(Particle filter). TBD, RCS ( SNR). TBD(Recursive TBD),,. (), --.,. Abstract This paper describes the technique for Radar Particle filter for TBD(Track Before Detect) processing. TBD technique is applied when target is difficult to detect due to low signal-to-noise ratio caused by strong clutter environments, small RCS targets and stealth targets. Particle filter is suitable for a recursive TBD algorithm and has improved estimation accuracy than Kalman filter. In this paper, we will present a new method of calculating particle weight, when observation values(including strong clutter) are received at the same time. Estimation error performance of the particle filter algorithm is analyzed by using the virtual radar observation scenario. Key words: Radar Tracking Filter, Particle Filter, TBD, Track-Before-Detect, Stealthy Target. 서론 --(DBT: Detect Before Track).,, CFAR., --(TBD: Track Before Detect), RCS ( SNR) [1] [4]. 90 % 10 6. SNR. DBT RCS, [4],[10]. 2015(). (RadarEW R&D Center, Hanwha THALES) *(Graduate School of Convergence Science and Technology, Seoul National University) Manuscript received January 26, 2016 ; Revised February 26, 2016 ; Accepted March 7, 2016. (ID No. 20160126-011) Corresponding Author: No-Jun Kwak (e-mail: nojunk@snu.ac.kr) c Copyright The Korean Institute of Electromagnetic Engineering and Science. All Rights Reserved. 317
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. vol. 27, no. 3, Mar. 2016. TBD, SNR / [4],[10]. TBD (dynamic programming) (particle filter) [5],[8],[10]. Viterbi. [5][7], (batch method) [8],[10]. (particle filter) [5],[8]. (particle filter), TBD(Recursive TBD) [4],[8],[10]. (Kalman filter),. RCS,. TBD. 2, (),. 3,. 4.. 파티클필터 2-1 시스템상태방정식 (System State Model) (1)., x, y, z, x, y, z, k. (1) (2), A W (3). A, W. (2) (3) (3) T, W σ 2. 2-2 관측모델 (Measurement Model) (,, ), (4). h() s k, n k z k [11]. (4) N (6), [11]. (5) (6) N, (7) [11]. (7)., N, 318
TBD (), ( RCS). N(), (8).. CFAR, ( + ), 3. --, (EKF).,.,,. 1., 1. (A) (B) (9), (10) (11)., PD SNR P fa. k m, m PD m (9), m w m (10). Q( ) Marcum s Q-function. r m, v m. w m w={w 1, w 2,... w m }, (11). log (9) (8) 표 1. Table 1. Characteristics of tracking methods applied in this paper. (A) (B) (C) 그림 1. Fig. 1. The proposed weight calculation method. (10) (11). 1,. k m N (12), (13), (14) (15). log (12) 319
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. vol. 27, no. 3, Mar. 2016. (13) (14) (15) m i (13). m N (14), (15). 2-3 리샘플링 (Re-sampling) (Re-sampling),. N eff, N th. (16). (δ) (17)., (18). (16) (17) (18) (18) N, (8). (2).. 시뮬레이션 3-1 레이더시뮬레이션시나리오설정 2 그림 2. Fig. 2. Moving path scenario of low-intercept target (low SNR & low RCS target). 표 2. Table 2. Radar system parameter for simulation. System parameter Operating frequency. f L-band Transmit peak power P t - Pulse width Τ - N(# of pulse) N 100 Tx antenna gain G t 35 dbi Rx antenna gain G r 35 dbi System noise figure F 10 db RCS Front σ front Ref. dbsm Top σ top (Ref.+5) dbsm Side σ side (Ref.+3) dbsm. x z, y. 2. RCS σ front, σ top, σ side, RCS, SNR. 3.. 4. 320
TBD 처리를 위한 레이더용 파티클 필터 기법 연구 실제 이동경로 (a) (a) Real path (b) SNR=12 db (c) (b) Measurement path (c) (SNR Threshold level=12 db) SNR=9 db Measurement path (SNR Threshold level=9 db) 이동 경로에 따른 그림 3. SNR Fig. 3. SNR according to the simulation scenario. (d) SNR=7 db (e) (d) Measurement path(e) (SNR Threshold level=7 db) SNR=5 db (f) SNR=3 db Measurement path (f) Measurement path (SNR Threshold le(snr Threshold level=5 db) vel=3 db) 임계치에 따른 관측경로 및 클러터 그림 5. SNR Fig. 5. Measurement path with clutter according to SNR threshold level. 레이더 클러터를 포함한 시뮬레이션 시나리오 그림 4. Fig. 4. Simulation scenario including radar clutter. 3-2 임계치 설정에 따른 관측값 임계치 설정에 따른 관측경로 및 클러터의 변화를 그 림 5에 보인다. 임계치가 높을수록 클러터를 효과적으로 제거할 수 있으나, 표적의 SNR이 낮은 구간에서 관측값 이 존재하지 않는다. 반대로, 임계치를 낮추면 복잡한 클 러터까지 수신되기 때문에, 선형화된 칼만필터 적용에 한 계가 있다. 3-3 관측값 처리 범위 설정 임계치를 낮추면 특정한 시각에 유입되는 관측값 이 매우 많고, 특히 전체 공간상에 관련 없는 클러터 값들 이 존재한다. 따라서 추정된 값을 기준으로 반경 R 이내 의 값만을 처리토록 구현한다. 이때 반경 R은 타겟의 이 동속도를 고려하여 설정한다. 이를 그림 6에 보인다. SNR 효율적인 처리를 위한 관측값 처리 범위 제한 그림 6. Fig. 6. Constraints of processing measurement values for improving efficiency. 3-4 시뮬레이션 그림 4에서 보인 시뮬레이션 시나리오를 가지고, 임계 치 SNR 고정 5 db 이하를 제거하고, 관측값 처리 범위를 각각 2.5 km, 1 km로 제한한 후에 수신된 관측값(클러터 포함)을 그림 7에 보인다. 321
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. vol. 27, no. 3, Mar. 2016. (a) 2.5 km (a) Limitation R=2.5 km (b) 1 km (b) Limitation R=1 km 그림 7., Fig. 7. Measurement values after limiting processing area. 그림 9. ()() : 600 Fig. 9. Real path vs estimated path : particle number 600. (a) 2.5 km (a) Limitation R=2.5 km (b) 1 km (b) Limitation R=1 km 그림 8. ()() : 100 Fig. 8. Real path vs estimated path : Particle number 100. 7 8., 2.5 km. 3-4... 9 600. 7,. 3-5 추정성능분석 RMSE(Root Mean Square Error). (19) (20). (19) k n. (20) k. M error, M pos, R pos, E error, E pos. (19) (20) 10( 600)., R. SNR. E error,avg (21), 11 12. 20. (21) 322
TBD. 3-6 기존방식 ( 칼만및파티클필터 ) 들과의성능비교 그림 10. (: 600) Fig. 10. Measurement error vs estimation error. 1 3. SNR 3, 13 3. (C), SNR,., (A) 3 그림 11. Fig. 11. Estimation error according to the number of particles. 그림 13. Fig. 13. Estimation error according to the tracking algorithm. 그림 12. Fig. 12. Average estimation error according to the number of particles. 12. 600 표 3. Table 3. Average estimation error according to the tracking algorithm. (k=1100) () (A) 332 m 236 m (B) 291 m 214 m (C) () k time index. 221 m 124 m 323
THE JOURNAL OF KOREAN INSTITUTE OF ELECTROMAGNETIC ENGINEERING AND SCIENCE. vol. 27, no. 3, Mar. 2016.. SNR (C) 2.,,.. 결론 ( + ), --, TBD., TBD 2. References [1] D. J. Salmond, H. Birch, "A particle filter for track-before-detect", Proceedings of the American Control Conference, vol. 5, pp. 3755-3760, 2001. [2] Bocquel, Mélanie, Hans Driessen, and Arun Bagchi, "Multitarget particle filter addressing ambiguous radar data in TBD", Radar Conference(RADAR), pp. 575-580, 2012. [3] Zhaoping Wu, Tao Su, "Radar target detect using particle filter", Radar Conference 2010 IEEE, pp. 955-958, May 2010. [4] Y. Boers, J. N. Driessen, "Multitarget particle filter track before detect application", IEE Proceedings-Radar, Sonar and Navigation, pp. 351-357. Dec. 2004. [5] S. Tugac, M. Efe, "Radar target detection using hidden markov models", Progress in Electromagnetics Research B, vol. 44, 241-259, 2012. [6] L. A. Johnston, V. Krishnamurthy, "Performance analysis of a dynamic programming track before detect algorithm", IEEE Transaction on Aerospace and Electronic Systems, vol. 38, pp. 228-242, 2002. [7] Shane M. Tonissen, Robin J. Evans, "Performance of dynamic programming techniques for track-before-detect", Aerospace and Electronic Systems, IEEE Transactions on 32.4, pp. 1440-1451, 1996. [8] Mark G. Rutten, Neil J. Gordon, and Simon Maskell, "Recursive track-before-detect with target amplitude fluctuations", IEE Proceedings-Radar, Sonar and Navigation, pp. 345-352, 2005. [9] Mark G. Rutten, Branko Ristic, and Neil J. Gordon, "A comparison of particle filters for recursive track-beforedetect", Information Fusion, 2005 8th International Conference on, vol. 1, pp. 169-175, 2005. [10] Biruk K. Habtemariam,, Ratnasingham Tharmarasa, and Thia Kirubarajan, "PHD filter based track-before-detect for MIMO radars", Signal Processing, pp. 667-678, 2012. [11] Y. Boers, J. N. Driessen, "Particle filter based detection for tracking", American Control Conference, vol. 6, pp. 4393-4379, 2001. 2004 8: () 2007 2: () 2015: () 2007: [ 주관심분야 ],, / 1999 2: ( ) 2001 2: ( ) 2007 3: [ 주관심분야 ], 324
TBD 1997 2: () 1999 2: () 2003 2: () 2003 32006 8: 2006 92007 2: BK 2007 32013 8: / 2013 9: [ 주관심분야 ],,, 325