Robust visual tracking based on improved perceptual hashing for robot vision

Abstract

In this paper, perceptual hash codes are adopted as appearance models of objects for visual tracking. Based on three existing basic perceptual hashing techniques, we propose Laplace-based hash (LHash) and Laplace-based difference hash (LDHash) to efficiently and robustly track objects in challenging video sequences. By qualitative and quantitative comparison with previous representative tracking methods such as mean-shift and compressive tracking, experimental results show perceptual hashing-based tracking outperforms and the newly proposed two algorithms perform the best under various challenging environments in terms of efficiency, accuracy and robustness. Especially, they can overcome severe challenges such as illumination changes, motion blur and pose variation.