Cooperation of acoustic and vision data for multitarget tracking

This paper explains the implementation of a real time bi-target tracking using both acoustic and video data. We show how the cooperation of highly heterogeneous sensors may improve the overall efficiency. These data are filtered using Kalman filtering techniques. We send reconfiguration orders to the actuators for an optimal new data sampling. The observed scene raises some classical control problems like the partial observability conditions, the unpredictable behaviors for the different components of the world because of the limited a priori knowledge and the synchronisation of the data