Performance and complexity analysis of adaptive particle filtering for tracking applications

This paper provides a performance and complexity analysis of particle filtering as applied to real-time object tracking. The number of particles and the sampling rate influences the performance of particle filters, but more importantly, they affect very much their complexity. In this paper, we propose a particle filter that changes the number of used particles during filtering, where the number of particles is employed for making decisions about performing resampling. The performance of the proposed particle filters is demonstrated on the bearings-only tracking problem.