Target Motion Analysis via Hard and Soft Data Fusion

Abstract

Target Motion Analysis (TMA) requires the online fusion of multiple hard and soft data sources for target tracking. This paper proposes a Bayesian filtering solution for multisource fusion with hard and soft data. Appropriate models for various types of hard and soft data are developed so that they can be fused in a consistent manner under the Bayesian framework. The resulting Bayes filter is highly non-linear and non-Gaussian. Hence, a parallel particle filter is developed to facilitate a user adjustable trade-off between computation time and tracking accuracy. Numerical studies on realistic scenarios are also presented.