Convoy detection processing by using the hybrid algorithm (GMCPHD/VS-IMMC-MHT) and Dynamic Bayesian Networks

Convoys are military objects of interests in certain applications like battlefield surveillance, that is why it is important to detect and track them in the midst of civilian traffic as part of the situation assessment. Our purpose is a process in two steps. The first is an original tracking algorithm appropriate for ground moving target indicator (GMTI) data based on the hybridization of a labeled GMCPHD (Gaussian mixture cardinalized probability hypothesis density) and the VS-IMMC-MHT (variable structure - interacting multiple model with constraints - multiple hypothesis tracking): one is very efficient to estimate the number of targets and the other for the state estimates. Then, by using algorithm outputs and other data like video or SAR if they are available, vehicle aggregates are detected and their characteristic are introduced into a dynamic Bayesian network which processes the probability for an aggregate to be a convoy. Finally, the number of targets belonging to the convoy is evaluated. This process is tested on a complex simulated scenario, our tracking algorithm is compared to classical ones and used to compute the probability to have convoys.