Real-Time Correlation Filter Tracking by Efficient Dense Belief Propagation with Structure Preserving.

Rui Yao,Shixiong Xia,Zhen Zhang,Yanning Zhang

DOI: https://doi.org/10.1109/tmm.2016.2631727

IF: 7.3

2017-01-01

IEEE Transactions on Multimedia

Abstract:Patch-based models that combine local image features or regions into loose geometric assemblies are a powerful paradigm for visual object tracking, and they present favorable properties such as robustness to partial occlusion, deformation, and the ability to address viewpoint changes. However, effectively exploiting the spatial-temporal confidence scores of each patch to construct a robust tracker while ensuring a low computational cost with dense discrete search remains a challenging problem. In this paper, we propose a unified Markov random field (MRF) model that can effectively capture spatio-temporal intrapatch relations and occlusion priors to enhance the tracking performance, and we derive a highly efficient dense belief propagation for inference of the proposed MRF model. We propose a tracker that models the tracking object with a constellation topology (i.e., a global object and several local patches), where the graph structure model describes the pairwise spatial structure and the image observation model corresponding to the correlation filter with occlusion handlingmeasuring the appearance similarity. Furthermore, these two models are updated online by exploiting the flexibility of local patches. Extensive experimental results on single-and multipleobject tracking show that the proposed algorithm performs favorably against state-of-the-art methods and runs in real time.