IET Computer Vision (Jun 2014)

Tracking with scattering descriptor

  • Xiaolin Tian,
  • Licheng Jiao,
  • Xiaowei Shang

DOI
https://doi.org/10.1049/iet-cvi.2013.0124
Journal volume & issue
Vol. 8, no. 3
pp. 195 – 206

Abstract

Read online

This study proposes a new method to track the moving object based on undecimated scattering transform (UST). The UST removes the down‐sampling operation from the traditional scattering transform to produce a complete representation. Based on the UST, the structural information of object can be captured, which is a highly discriminative representation and facilitates the tracker to distinguish the object from background. The update parameter of tracking model is adaptively adjusted by the correlation coefficient. Occlusion identification is achieved by using a new cascaded model, which can correctly find occlusion and avoid the drift. Experimental results demonstrate that the proposed method is able to track the object accurately and reliably in realistic videos where the appearance and motion change drastically over time.

Keywords