Tracking multiple feature in infrared image with mean-shift

Abstract

Mean-shift algorithm with robust performance is one of the well-known tracking algorithms. Tracking targets with Mean-shift algorithm is tracking the statistical features of their pixels by the histograms. The classic Mean-shift for tracking targets based other features has not been developed. In this paper, we propose a strategy which can make Mean-shift track multiple types of features of targets. We first map the features into the pixel intensity and create the feature images. Then these feature images are used to construct multiple feature images (MFIs). The kernel density estimation algorithm tracks targets in MFIs can indirectly track these features. © 2011 Springer-Verlag.