Tracking Multiple Moving Objects from Monocular Image Sequences with Hausdorff Distance and Context Information

Abstract

This work investigates a new approach to the tracking of multiple moving objects. The method comprises of two level procedures. At the first level, it tracks each object by comparing its shapes between successive frames using Hausdorff distance. At the second level, it uses context information to reduce the complexity of the problem and break it into several more manageable ones. It means that the condition about each object (in this case: correspondence, state, etc.) will be evaluated at each frame and the proper tracking procedure will be initiated based on its evaluation at each object basis. The main contribution of our work is the development of searching and updating procedures and also objects management procedure which makes the system capable to track objects under various, real condition. Experiments conducted on real images demonstrate that the approach is robust against overlapping and various objects' condition.