Nonparametric techniques for graphical model-based target tracking in collaborative sensor groups

Target tracking using collaborative sensor groups is an effective mechanism for reducing the scalability issues in distributed sensor networks. Using graphical models for such a sensor group together with appropriate class of nonparametric message passing algorithms, we explore efficient approaches to handle the related data fusion problems characterized by spatially distributed observations. Messages consisting of multiple Gaussian components have been efficiently handled with the help of nonparametric belief propagation techniques. The advantage of such an approach in a myopic radar network has been verified here using Monte Carlo simulations by comparing the tracking performance obtained with centralized and distributed fusion schemes.