Improvement of track segment association with classification information

Benjamin Pannetier (1) , Jean Dezert (1)

1 ONERA - The French Aerospace Lab [Palaiseau]

In this paper, we propose a full process to track ground targets and correct the ground tactical situation. The ground target tracking is done at each scan time of a Ground Moving Target Indicator (GMTI) airborne sensor. The algorithm takes into account the road network in the tracking process to improve the track precision and an Interacting Multiple Model (IMM) to deal with maneuvers of the targets. The multiple target tracking algorithm based on Structured-Branching Multiple Hypothesis Tracker (SB-MHT) improves significantly the track precision and continuity. However, because of subversive target maneuvers (a target can deliberately stop to avoid to be detected by GMTI sensor), the airborne sensor maneuver who carries along the cut-off of the sensor, or because of a possible poor road network modelling, the obtained tracks are not always properly updated and can be automatically and erroneously deleted by classical algorithms. To circumvent this serious problem, we propose an extension of our previous work to correct the current situation with past situation in order to correlate current tracks with past tracks and solve the track segment association problem. The correlation between current tracks with old ones is based on kinematic and classification information. The performances of this global process are quantified on a simulated scenario considering twenty maneuvering ground targets observed by one airborne with a GMTI sensor and Unattended Ground Sensor (UGS).

Mots clés: Ground target tracking; track segment association

Domaines: Sciences de l'ingénieur [physics]

Benjamin Pannetier, Jean Dezert. Improvement of track segment association with classification information. 2014. ⟨hal-01080986⟩