Robust object tracking using motion context in crowded scenes

Abstract

Tracking objects in a crowded scene with occlusions has been a challenge in computer vision and multimedia in the past years. This paper presents a novel framework to track any arbitrary object through modeling its coupled motion context. For a scene which is densely packed, an individual movement is restricted into a specific pattern to make it regular to be detected. Moreover, members in a crowded scene are clustered into groups and thereby modeled as crowded scene contexts of a tracking object. Accordingly, we present a novel framework to track motions for any object in a crowded scene even with occlusions by using the modeled contexts. Experiments on a number of real-life surveillance videos illustrate the effectiveness and robustness of our method especially in handling occlusions in crowded scenes. © Springer International Publishing Switzerland 2013.