need detailed mathematical description(books/tutorial/papers) for Bayesian net applied for object tracking applications.
Hi,
you can find a full description of bayesian networks in the Phd Thesis of Murphy : "Dynamic bayesian networks: representation, inference and learning" (2002).
Another paper is the one published by Nillius "Multi-Target Tracking - Linking Identities using Bayesian Network Inference" , IEEE Computer Society Conference on Computer Vision and Pattern Recognition (2006).
What kind of tracking do you study ? (camera, radar,... ? )
Hello. As far as I know, Bayes nets are not used for object tracking, per se. The idea of a Bayes net is to represent conditional probability relationships (dependencies) in a multivariable stochastic system. A Bayes net might be a compoent of an overall tracking system where there is a complicated probabilistic relationship betwen the sensor and the state of the object to be tracked. In the systems that I am aware of, the modelling of this relationship does not require Bayes nets since there is an obvious physical relationship based on optics, electromagnetics, etc. Bayes nets tend to crop up more in systems where the physics is not known, like cause and effect relationships. I don't think that Bayes nets usually allow the variables to change in a dynamical relationship to each other, but I could be wrong.
On the other hand, Bayesian inference in the sense of probability theory is almost universally used in target tracking, where there is a definite dynamical relationship between the underlying stochastic variables. Please take a look at the review paper in the link below for examples of such techniques.
Article Taxonomy of multiple target tracking methods
- Badges
- Science method