Survey of Visual Tracking Algorithms in the Complex Scenarios

Abstract

Visual object tracking is a fundamental problem in computer vision. It has been widely used in civilian and military fields, such as battlefield reconnaissance, video surveillance, automatic driving, video analysis, and many other areas. In recent years, although the object tracking algorithm has made great progress, stable object tracking is still a challenging task due to random target changes and complex scenarios. Firstly, the difficulties and challenges in actual tracking scenarios are introduced in this paper. Then, aiming at the background clutter, rotation changes, occlusion, and scale changes, the representative discriminative object tracking algorithms are summarized and analyzed from the perspective of feature extraction, observation model, and model update mechanism. Subsequently, 25 typical tracking algorithms are evaluated and analyzed on OTB2015 database. Finally, the further research directions are prospected.

Keywords

• |visual object tracking|discriminative visual tracking model|correlation filter|deep learning