Sequential Bayesian inference models for multiple object classification

This paper explores the use of a Bayesian inference model for updating the classifications of multiple objects simultaneously when given a measurement on only one of the objects. It is proven that with prior knowledge on the number of objects being tracked, each measurement can update the probability mass function for every tracked object. This result is generalized to sensors that can only classify subsets of the objects. The paper also shows empirically that the rate of convergence to the correct classifications for all objects using this model is improved over tracking each object independently. Finally, the paper ends by demonstrating the efficacy of the model by fusing measurements from two different classification sensors in a multiple target tracking scenario.