In track to track correlation and fusion, the tracks are first correlated together, typically using one of the following approaches-greedy nearest neighbor, global nearest neighbor, or Murty's K-best hypothesis. Once the correlation relationships have been determined, then the tracks are fused together to improve the state estimate and covariance of the track representing each object. In the case of Murty's K-best approach, the correlation decisions from the best hypothesis are used to determine which tracks are then fused together. In this paper, we propose using a blending fusion approach using the correlation results from a set of the highest scoring hypotheses. This approach “hedges the bets” by using more than one hypothesis thus increasing the probability that the resulting covariance will contain the true target even in the case when the best hypothesis includes a miscorrelation. The result is that the blended fused tracks will in most cases have a larger covariance than the fused tracks from the best hypothesis with the tradeoff being a higher covariance containment probability. This paper builds upon the previous work by Ogle, et al. using a multiple level fusion system to investigate the effects of using blended fusion. Monte Carlo simulations were performed, and pattern metrics were used to assess the effectiveness.
Multi-Target Blended Track Fusion in the Presence of Sensor Biases
2020-03-01
643328 byte
Conference paper
Electronic Resource
English
CORRESPONDENCE - Track Association and Track Fusion with Nondeterministic Target Dynamics
Online Contents | 2002
|