Tracking Completion
A fundamental component of modern trackers is an online learned tracking model, which is typically modeled either globally or locally. The two kinds of models perform differently in terms of effectiveness and robustness under different challenging situati
- PDF / 3,853,845 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 7 Downloads / 192 Views
Department of EECS, University of Kansas, Lawrence, KS 66045, USA [email protected], [email protected] 2 China Unicom Research Institute, Beijing 100032, China [email protected] 3 Department of EE, Tsinghua University, Beijing 100084, China [email protected] 4 National Laboratory of Pattern Recognition, Institute of Automation, CAS, Beijing, China
Abstract. A fundamental component of modern trackers is an online learned tracking model, which is typically modeled either globally or locally. The two kinds of models perform differently in terms of effectiveness and robustness under different challenging situations. This work exploits the advantages of both models. A subspace model, from a global perspective, is learned from previously obtained targets via rank-minimization to address the tracking, and a pixel-level local observation is leveraged simultaneously, from a local point of view, to augment the subspace model. A matrix completion method is employed to integrate the two models. Unlike previous tracking methods, which locate the target among all fully observed target candidates, the proposed approach first estimates an expected target via the matrix completion through partially observed target candidates, and then, identifies the target according to the estimation accuracy with respect to the target candidates. Specifically, the tracking is formulated as a problem of target appearance estimation. Extensive experiments on various challenging video sequences verify the effectiveness of the proposed approach and demonstrate that the proposed tracker outperforms other popular state-of-the-art trackers. Keywords: Matrix completion · Object tracking Local observation · Appearance estimation
1
·
Subspace model
·
Introduction
Visual tracking is an important topic in computer vision for its various applications, such as video analysis, robotics, and visual surveillance. In general, tracking models can be mainly classified into two categories: global and local. Global model exploits the overall information that varies in the entire target region. Local model treats the target as a series of small image patches to focus on the changes in each small region. It has been demonstrated that the global model is robust to some holistic appearance changes, like illumination variations and c Springer International Publishing AG 2016 B. Leibe et al. (Eds.): ECCV 2016, Part VIII, LNCS 9912, pp. 194–209, 2016. DOI: 10.1007/978-3-319-46484-8 12
Tracking Completion
195
pose changes [1–4]. The local model, on the other hand, is intrinsically effective to the challenges, such as partial occlusions and local deformations [5–8]. This is because only some of the local patches are influenced by the distractive objects (noise contaminated regions), while the rest are considered to be noise-free. To effectively deal with various appearance changes, a robust tracker is desired to be able to exploit the advantages of both global and local tracking models. In this work, we propose to leverage the effectiveness of the global method in
Data Loading...
