Student research highlight: Gyro-aided visual tracking using iterative earth mover's distance

Abstract

Visual tracking is an important problem in the field of robotics, especially for the cameras mounted on agile flying robots. The information generated from these images by the tracking algorithm can be used by autonomous vehicle navigation [1], or human-robot interaction [2], [3]. Image-based tracking algorithms are categorized as point tracking, kernel tracking, or silhouette tracking [4]. First, distinguishing features, such as color, shape, and region are selected to identify objects for visual tracking. Then, the tracked object is modeled, on the basis of the selected features. The correspondence or similarity measurement between the target and the candidate across frames is constructed, such as the sum of squared differences of the pixel intensity values [5], the mutual information (MI) [6], [7], the normalized cross correlation, and the Bhattacharyya coefficient [8]. Supervised or unsupervised online learning algorithms are also used for visual tracking. In this article, a new visual tracking algorithm called iterative EMD (iEMD), on the basis of the optimization of the EMD between templates, is presented. A preliminary version of this work was recently published [18].