Multi-object tracking and identification via particle filtering over sets

Abstract

The ability for an information-fusion system to track and identify potentially multiple objects in a dynamic environment is essential for many applications, such as automated surveillance, traffic monitoring, human-robot interaction, etc. The main challenge is due to the noisy and incomplete perception including inevitable false negative and false positive errors, usually originated from some low-level sensors or detectors. To address this challenge, we propose a novel particle filtering over sets based approach to multi-object tracking and identification. We model the multi-object tracking problem as a hidden Markov model with states and observations represented as finite sets. We then develop motion and observation functions accordingly, and do the inference via particle filtering. The corresponding object identification problem is then formulated and solved by using the expectation-maximization method. The set formulation enables us to avoid directly performing observation-to-object association. We empirically confirm that the proposed algorithm outperforms the state-of-the-art in a popular PETS dataset.