Modelling of the background (“uninteresting parts of the scene”), and of the foreground, play important roles in the tasks of visual detection and tracking of objects. This paper presents an effective and adaptive background modelling method for detectin
The consensus based background modelling and consensus based appearance modelling
are combined to form an effective tracking system.
Experiments are presented to show that the proposed methods can achieve promising performance in background subtraction (including handling both static and dynamic background scenes) and foreground appearance modelling (including tracking/segmenting people through occlusions).
The organization of the remainder of this paper is as follows: in section 2, we first present a short review of previous related work on background modelling. Then, we present the SACON concept and a framework for applying SACON to background subtraction. In section 3, experiments showing the advantages of our background modelling method over several popular methods are provided. We also investigated the influence of the parameters of SACON on the results. In section 4, we describe how to use sample consensus to model the foreground appearance. Experimental results of segmenting and tracking people with occlusions by the proposed tracking method are also provided. We conclude the paper in section 5.
Part I: Background Modelling
2. Sample Consensus in Background Modelling – SACON
In this section, we define a novel Sample Consensus (SACON) method for modelling background scenario, and a framework that employs SACON as a core for background subtraction. We begin with an overview of related work on background modelling (section 2.1). We then introduce the concept behind SACON (section 2.2) described for conceptual simplicity in terms of RGB colour space. Various modifications to the concept are then introduced (section 4