Online Visual Object Tracking Using Incremental Discriminative Color Learning
Abstract
This paper presents a method for tracking an object in a sequence of images given its location in the first frame. Recently, a class of techniques called discriminative methods has shown promising results. These methods are based on training a classifier to distinguish the object from surrounding background. However, discriminative methods do not explicitly model the object. Therefore, noisy samples are likely to interfere and cause visual drift. In this paper, 3D joint RGB histograms of the object and surrounding background are used to develop an object model. An incremental color learning scheme with a forgetting factor is applied to evolve the object model during tracking. It is shown the proposed method can handle visual drift effectively. Evaluated against five state of the art methods, experiments demonstrate superior results of the proposed tracking algorithm. Implemented in MATLAB, the algorithm runs at 17.2 frames per second, including image input/output time.
Keywords
Visual Object Tracking, 3D Joint RGB Histogram, Log-Likelihood Ratio, Incremental Learning, Mean-Shift Localization
