Skip to Main Content
The utilisation of a network of heterogeneous sensors to track humans and analyse their behaviours in indoor environment is essential due to the high risk of occlusions. For this purpose, the Bayesian Occupancy (BOF) filter was shown efficient to fuse data coming from infrared and visible cameras by providing the occupancy/velocity probability distributions of each spatial cell of the grid representation of the environment. As the main contribution of this paper, we will present a novel generative sensor model intended to be used for 3D sensors providing range information (e.g., time-of-flight cameras). In order to show the effectiveness of our solution, we will present a fusion example using (i) two visible cameras, (ii) one infrared camera, (ii) and one PMD sensor. We will especially show that this fusion scheme significantly increase the robustness of the tracking process.