Real-Time Multi-Object Tracking using Random Finite Sets

The multi-object Bayes (MOB) filter uses random finite sets (RFSs) to represent a scene. A drawback of this filter is the computational complexity of the multi-object likelihood function. In this contribution, an approximation of the multi-object likelihood function is presented allowing for real-time implementation on a graphics processing unit using sequential Monte Carlo (SMC) methods. Additionally, a track extraction algorithm using clustering as well as an approach to determine the existence probability of each single object are proposed.