Hard negative mining for correlation filters in visual tracking

Visual tracking is a fundamental computer vision task. Recent years have seen many tracking methods based on correlation filters exhibiting excellent performance. The strength of these methods comes from their ability to efficiently learn changes of the target appearance over time. A fundamental drawback to these methods is that the background of the object is not modeled over time which results in suboptimal results. In this paper, we propose a robust tracking method in which a hard negative mining scheme is employed in each frame. In addition, a target verification strategy is developed by introducing a peak signal-to-noise ratio (PSNR) criterion. The proposed method achieves strong tracking results, while maintaining a real-time speed of 30 frame per second without further optimization. Extensive experiments over multiple tracking datasets show the superior accuracy of our tracker compared to state-of-the-art methods including those based on deep learning features.