A track-before-detect algorithm with thresholded observations
Transcript
A track-before-detect algorithm with thresholded observations
A track-before-detect algorithm with thresholded observations and closely-spaced targets Emanuele Grossi Marco Lops Luca Venturino Dipartimento di Ingegneria Elettrica e dell’Informazione Università degli Studi di Cassino e del Lazio Meridionale {e.grossi, lops, l.venturino}@unicas.it GTTI, Ancona, 2013 Multi-frame detection: two-stage approach raw data from the sensor traditonal detector list of candidate detections TBD processor to the tracker γ2 γ1 buffer (L − 1 frames) The traditional detector operates on a scan-by-scan basis and provides a list of candidate detections The primary threshold γ1 is set lower than that used in traditional single-scan detection, causing an increment of both PD and FAR The TBD processor exploits data correlation over multiple frames to confirm or delete each candidate detection The goal is to restore FAR to an acceptable level while maintaining part (if not all) of the detection gain 2/6 TBD processor 1 L−1 2 L space (n − L + 1)T time (n − L + 2)T nT a trajectory cones of physically admissible trajectories observation surviving the first stage observation censored by the first stage 3/6 TBD processor Step 1 Step 2 Estimated trajectories false trajectory false trajectory ctory tra je target ory ct tra je target SDR ≷ γ2 SDR ≷ γ2 false plot linked to the target trajectory jectory target tra target trajectory target tra jec tory time SDR ≷ γ2 true plot false plot time estimated trajectory space plot of the weak target linked to the trajectory of the strong target space space target tra jec tory time 1) Compute the best trajectory for each plot as in [1] 2) Extract the dominant trajectories and remove their plots 3) Iterate steps 1-2 until all trajectories are extracted 4) Compare the SDR of estimated trajectories against γ2 4/6 Experimental Setup radar parameters range azimuth range accuracy azimuth accuracy scan duration FARin FARout 40 to 140 km −60◦ to 60◦ 20 m 0.5◦ 1s 1000 per min 1 per min target parameters number initial range initial azimuth direction velocity 4 triplets random in [50, 130] km random in [−50◦ , 50◦ ] random random in [0, 300] m/s 5/6 PD and RMSE 1 1 d = 1 km Targets 2 and 3 0.8 SDR = 15 dB 0.8 Target 1, SDR = 12 dB PD 0.6 PD 0.6 0.4 0.4 0.2 0.2 0 0 900 900 850 SDR = 9 dB 850 all SDRs 800 750 750 RMSE [m] RMSE [m] All targets, all SDRs 800 700 650 Target 1, SDR = 12 dB SDR = 9 dB 700 650 Targets 2 and 3 600 550 500 600 L=1 L = 10, algorithm in [1] L = 10, proposed algorithm 5 10 15 20 SDR of Targets 2 and 3 [dB] 25 500 0 SDR = 15 dB L=1 L = 10, algorithm in [1] L = 10, proposed algorithm 550 1 2 3 4 d [km] 5 6 7 8 6/6