יום רביעי, 16 בדצמבר 2009

The Gaussian Mixture MCMC Particle Algorithm for Dynamic Cluster Tracking

אבישי כרמי


Signal Processing and Communications Laboratory
Department of Engineering, University of Cambridge, UK

The problem of multiple object tracking (MOT) poses major challenges for researchers in the fields of estimation and sensor fusion. Essential difficulties that are frequently encountered in MOT refer to a) data association, namely, the need to efficiently associate observations with their emitting sources in a highly cluttered environment, b) complex behavioral dynamics which might involve social interactions between objects, and c) adequate statistical representation of the overall scenery. In this talk I will elaborate on some of these issues and present a unified approach for tracking multiple clusters of coordinated objects based on a novel Markov chain Monte-Carlo (MCMC) filtering scheme.

Belonging to the family of sequential Monte-Carlo (SMC) algorithms, the newly proposed filter relies on a discrete approximation of the posterior probability density function of the joint state. As opposed to the conventional SMC approaches which tend to become inefficient in high-dimensional settings, the proposed filtering method naturally copes with complex scenarios consisting of many objects owing to its Metropolis-Hastings core. A demonstration of the new filter’s performance when applied for feature tracking in a video sequence will be presented.

The seminar will take place on Monday, 28.12.2009,
at 13:30, in the Seminar Room,
Asher Space Research Institute

Light Refreshments will be served before the seminar.

אין תגובות: