Visual tracking via perceptual image hash from a mobile robot

Abstract

Visual tracking is a typical issue for a mobile robot. In this paper, we propose a robust tracking algorithm which exploits perceptual image hash vector based on Discrete Cosine Transform (DCT) perceptual image hashing algorithms. We use perceptual hash vector of the target object and perceptual hash vectors surrounding the target object in a scene. The learned perceptual hash vectors are used to determine the similarity of all searching regions to update for the next frame. Tracking in the next frame is formulated by computing a hamming distance as a problem that integrates the searching local context information. And the best object location can be estimated by minimizing the hamming distance. Our proposed algorithm is robust to appearance variations introduced by illumination changes, scale and pose variations.