Discriminative Object Tracking Via Subspace Representation with Superpixel Dimension Reduction

Abstract

Visual object tracking is a core research area in the eld of pattern recognition and newlinecomputer vision. It is the process of tracking an object of interest from frame to newlineframe in an image sequence. It can be applied in various applications of computer newlinevision like video surveillance, robot navigation, trac monitoring, vehicle tracking, newlinemedical system, activity recognition etc. But tracking of a visual object is not an newlineeasy task as it is always restricted by diculties caused by pose change, illumination newlinevariation, in-plane rotation, out-plane rotation, shape deformation, background newlinecluttered, motion blur, pose change and so on. Many object tracking methods newlinehave been developed in the past but still, those diculties remain a challenging newlineproblem. Though some tracking method can handle these diculties to some extend newlinebut their tracking speed and high computational complexity caused a problem. newlineThus object tracking remains as a demanding research work. newline newline