Context-based methods for track association

Target tracking systems use sensor information to estimate and predict current and future locations of target objects. An important question in target tracking is how much improvement in accuracy can come from using contextual information. Context information is often sparse and uncertain sensor information that decreases the accuracy of tracklet association. Context information, such as radio communicated heading and velocity may improve knowledge about future target locations or maneuvers. This investigation evaluates statistical methods for measuring the association between a radar track report and a contextual report from another sensor, in a Monte Carlo simulation. The performance of chi-squared statistics (such as the ´Mahalanobis Distance´), distributional distance (or ´Integrated Product´), and data imputation is measured in terms of association accuracy and speed.