Joint Data Association and Tracking for Surveillance Radar Using Gaussian Processes

Abstract

The growing presence of unmanned aerial vehicles or drones has increased the necessity of developing methods to detect and track them to secure restricted airspaces. This can be done using radars which is a common sensor around e.g. airports, an emerging method to track targets is by using Gaussian processes. However, false alarms is ever present in radars especially if tracking weak targets complicating the data association task. We propose a joint data association and tracking algorithm for a single target using a variational Bayesian approach to approximate the full posterior of both the association variable as well as the track. We show that the proposed algorithm outperforms a state of the art data association and tracking algorithm in the form of a Bernoulli filter, as well as the classical Kalman filter, for a large range of detection probabilities.