The probabilistic data association filter (PDAF) is a Kalman-based tracking approach to situations in which non-target (clutter) measurements may disguise the desired tar-get measurements. Throughout recent decades, feature-based modifications have been proposed to provide additional target-clutter discrimination beyond the proximity metric used in the original PDAF. Examples include amplitude information, directional deviation, and doppler deweighting, of which the first has become widely used under the title PDAFAI. This paper proposes a new feature for the PDAF, known as a speed deviation. It is based on the kinematics between a candidate measurement and the track, and aims to yield a more stable and effective mapping of measurements to target probabilities. A large collection of simulated target motion data (1,000 random-ized runs) is used to statistically demonstrate the effectiveness of this feature in reducing kinematic track error. Single-target motion is simulated here as a randomized series of constant velocity and constant rate of turn segments. The true target measurement is then corrupted by Gaussian noise, disguised in a uniformly distributed clutter field, and has the potential to “drop out” at any given time step. The target state is tracked using measurements of position and doppler recorded at regular intervals. Statistical analyses are carried out and demonstrate a significant improvement over the baseline PDAFAI framework. Kinematic error results are also shown across a wide range of target dropout rates, showing a consistent improvement over the baseline across various levels of tracking difficulty. Moreover, performance comparisons are made with respect to the more recent directional deviation and doppler deweighting features, against which the proposed speed deviation shows the single greatest error reduction.
A Speed Deviation Feature for Enhanced Target Tracking and Data Association
04.03.2023
2118063 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Feature-based probabilistic data association and tracking
Kraftfahrwesen | 2010
|Feature-based probabilistic data association and tracking
Tema Archiv | 2010
|Europäisches Patentamt | 2022
|Cross Entropy Data Association for Multiple Target Tracking
British Library Conference Proceedings | 2007
|Data association for target tracking by IR sensors
Online Contents | 2007
|