Detecting occlusion and camouflage during visual tracking

Abstract

Visual tracking is an important scientific problem; the human visual system is capable of tracking moving objects in a wide variety of situations. It is also of considerable practical importance; many actual and potential applications of visual tracking algorithms exist in domains such as surveillance, medicine, robotics and the media. Although many effective tracking algorithms exist, occlusion and camouflage remain a common problem. These can cause a tracker to become dissociated from its target, so that the data it produces is unrelated to the target's behaviour. We focus on the detection of occlusion and camouflage during particle filter-based tracking. We use a Gaussian Mixture Model of particle distribution, extracted via the EM algorithm, to investigate the effects of occlusion and camouflage on the particle set representing a given target. The information gained from this investigation informs the design of process-behaviour chart which alerts the tracker of the occurrence of occlusion or camouflage. Data produced by the process-behaviour chart is also used to map out the boundary of the interfering object, providing valuable information about the viewed environment.