Skip to Main Content
As evidenced by the works of many recent authors, the particle filtering (PF) framework has revolutionized probabilistic visual target tracking. In this paper, we present a new particle filter tracking algorithm that incorporates the switching multiple dynamic model and the technique of state partition with parallel filter banks. Traditionally, most tracking algorithms assume the target operates according to a single dynamic model. However, the single model assumption causes the tracker to become unstable, especially when the target has complex motions, and the camera has abrupt ego-motions. In our new tracking algorithm, the target is assumed to operate according to one dynamic model from a finite set of models. The switching process from one model to another is governed by a so-called jump Markov process. This strategy can effectively capture the target's dynamics. In addition, we have used the state partition technique and a parallel bank of extended Kalman filters (SP-PEKF) to generate the proposal distribution used in the particle filter to achieve further estimation accuracy. We have conducted the testing for the new tracking algorithm, and key outcomes are given in the results section. The preliminary result demonstrates that this new approach yields a significantly improved estimate of the state, enabling the new particle filter to effectively track human subjects in a video sequence where the standard condensation filter fails to maintain track lock.