Aggregation of Composition States for Markov Estimation in Level 2 Fusion
Stephen Stubberud, Kathleen Kramer
In sensor fusion, the use of composition information can help define and understand relationships between targets. This process, part of the Situational Assessment problem, also referred to as Level 2 fusion, can be quite complex when using standard classification approaches such as the Bayesian taxonomy. Determination of the number and type of elements that comprise a group can vary from report to report based on the type of sensors, the environment, and the behavior of the group. Estimation of group composition that can take these factors into account has been developed using a Markov chain approach. If the number of potential target classes is significant and the various standard group compositions are numerous, the computational complexity becomes unmanageable. This effort investigates a useful and computationally attainable Level 2 composition state estimate based upon the use of state aggregation.