Performances in multitarget tracking for convoy detection over real GMTI data

A convoy is defined as a group of vehicles traveling together for mutual support and protection. It constitutes an object of high military interest in the context of the situation assessment. However, it is a challenging task to track and evaluate because convoy targets are very close of each other. In this view, the Onera recently developed a new two step convoy detection process. The first is an original tracking algorithm appropriate for Ground Moving Target Indicator (GMTI) data based on the hybridization of two classical multitarget tracking algorithms, and adapted to closely spaced target tracking. Then, by using algorithm outputs and other data, vehicle aggregates are detected and their characteristics are introduced into a Dynamic Bayesian Network (DBN) which processes the probability for an aggregate to be a convoy. This process gives encourageous results with simulated data. In this paper, we validate this process by showing real data results.