Robust and Real-Time Object Tracking Using Scale-Adaptive Correlation Filters

Abstract

Correlation filter based tracking method has been widely used for its high efficiency and robustness. However, reducing model drifting while achieving both high robustness and fast scale estimation is still an open problem. In this paper, we represent the target in kernel feature space and train a classifier on a scale pyramid to achieve adaptive scale estimation. We then integrate three complementary features to further enhance the overall tracking performance. Extensive experiments have been conducted on the object tracking benchmark and the Princeton tracking benchmark. Experimental results show that our method achieves promising results on these benchmarks in terms of tracking accuracy, robustness and speed. It outperforms the state-of-the-art methods under nuisances of scale variation, illumination variation, deformation, in-plane rotation and out-of-plane rotation.