Tracking using Bayesian inference with a two-layer Graphical Model

Abstract

This paper introduces a new visual tracking technique combining particle filtering and Dynamic Bayesian Networks. The particle filter is utilized to robustly track an object in a video sequence and gain sets of descriptive object features. Dynamic Bayesian Networks use feature sequences to determine different motion patterns. A Graphical Model is introduced, which combines particle filter based tracking with Dynamic Bayesian Network-based classification. This unified framework allows for enhancing the tracking by adapting the dynamical model of the tracking process according to the classification results obtained from the Dynamic Bayesian Network. Therefore, the tracking step and classification step form a closed trackingclassification-tracking loop. In the first layer of the Graphical Model a particle filter is set up, whereas the second layer builds up the dynamical model of the particle filter based on the classification process of the Dynamic Bayesian Network.

Adaptive tracking algorithms to improve the use of computing resources

“…The approaches are very diverse. For instance, in [6], a two layer graphical model is used for tracking. In the first layer, a particle filter is set up, whereas the second www.ietdl.org IET Comput.…”

“…On the contrary, the operation of the classifier is often very fast and negligible compared with the load of the tracking algorithm. In comparison with other algorithms performing classification and tracking [6][7][8][9], our approach aims at reducing the computation time, by using a classifier to change in some way a base tracking algorithm. Another advantage of making an intelligent use of computation resources is that the system does not rely on special hardware or parallelisation.…”

“…The approaches are very diverse. For instance, in [6], a two layer graphical model is used for tracking. In the first layer, a particle filter is set up, whereas the second layer builds up the dynamical model of the particle filter based on the classification process of a dynamic Bayesian network.…”

A Bayesian Network for online evaluation of sparse features based multitarget tracking

“…The model is calculated and updated at each instant of time by considering the previous track outputs. Dynamic Bayesian Networks have been widely used for analyzing the evolution of temporal sequences [8] [9]. They allow one to model the conditional dependence within and across the time slot for set of dynamically changing variables.…”