Particle filtering using multiple cross-correlations for tracking occluded objects in cluttered scenes

This paper is concerned with the tracking of partially or entirely occluded objects in a video sequence. We propose certain modifications to the template matching approach, which seem to fit the type of tracking data being considered in the present note. Specifically, we will use a nonstandard particle filtering method via the following two steps: The first step employs the normalized cross-correlation function as the likelihood. The second step is to resample, and to fuse the results of multiple cross-correlations of different patches of the given object, in order to refine the likelihood for the particle filter. Experimental results show that the method is reliable for noisy measurements, and provides accurate results in cases of occlusion or heavy shadows.