A Robust Human-Machine Hybrid Visual Tracking System

Visual object tracking is a basic task in computer vision, which has been widely used in intelligent transportation, autonomous driving, security systems, military reconnaissance and other fields. Most studies in visual object tracking assume that the target changes smoothly and the target will not disappear for a long time. However, in practical applications, challenges such as complete occlusion, rapid movement and target appearance dramatic change, make it very difficult to track the target consistently for a long time. In this work, we propose a human-machine collaboration method to cope with such challenges. We hope to build a tracking framework that combines the powerful tracking capabilities of human vision with the state-of-the-art tracking methods, so as to get a robust visual tracking system. Humans participation can effectively improve the accuracy and robustness of tracking. In the process of human participation, the tracker can also improve its discrimination ability by recognizing the target of human interest. Compared with the state-of-the-art trackers, our method achieves higher performance on a fairly complex experimental dataset.