Online Multi-Object Tracking Based on Hierarchical Association Framework

Jaeyong Ju, Daehun Kim, Bonhwa Ku, David K. Han, Hanseok Ko; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2016, pp. 34-42

Online multi-object tracking is one of the crucial tasks in time-critical computer vision applications. In this paper, the problem of online multi-object tracking in complex scenes from a single, static, un-calibrated camera is addressed. In complex scenes, it is still challenging due to frequent and prolonged occlusions, abrupt motion change of objects, unreliable detections, and so on. To handle these difficulties, this paper proposes a four-stage hierarchical association framework based on online tracking-by-detection strategy. For this framework, tracks and detections are divided into several groups depending on several cues obtained from association results with the proposed track confidence. In each association stage, different sets of tracks and detections are associated to handle the following problems simultaneously: track generation, progressive trajectory construction, track drift and fragmentation. The experimental results show the robustness and effectiveness of the proposed method compared with other state-of-the-art methods.