Context Exploitation for Target Tracking

Abstract

Target tracking is the estimation of the state of one or multiple, usually moving, objects (targets) based on a time series of measurements. Widely addressed within the Bayesian statistical framework, it requires the modeling of the target state evolution and the measurement process. Information on the constraints posed by the context in which the target evolves and the measurement geometry is often available. This knowledge can be modeled, often in a statistical way, and integrated in the tracking filters to enhance their performance. This chapter presents several approaches to exploit different types of context knowledge and demonstrates context-enhanced tracking based on real and simulated data. Numerical results are given for the inclusion of sea-lanes in ship tracking and route propagation, and for road-map assisted air-to-ground radar tracking.