A multi-modal particle filter based motorcycle tracking system

Abstract

Object tracking in computer vision is an attractive research field due to its widespread application area and challenges. In the recent years, Particle filter is known as a prominent solution for the state estimation problems in point tracking and successfully applied in a wide range of applications. But one of its limitations is the weakness at constantly maintaining the multi-modal target distribution that may arise due to occlusion, clutter or the presence of multiple objects. Lately, that weak point has been overcome in a multi-modal Particle filter (MPF). This paper aims to build some most basic functions of a motorcycle surveillance system using MPF and basing on the color observation model. Accompanied with a simple but effective detecting strategy, the application has the processing rate equivalent to a real time tracking system and high performance. © 2008 Springer Berlin Heidelberg.