Multi-target tracking through occlusions using extended Kalman filter and network flows
Mengmeng Wang, Peixin Liu, Xiaofeng Li, Kai‐Da Xu · 2016
Multi-target tracking under the condition of occlusions is one of the difficult and attractive field in video tracking. Multi-target tracking has been modelled as network flows (NF) optimization problem recently. It is one of the most popular tracking-by-detection (TBD) algorithm, which links individual detections into trajectories. NF method is particularly effective due to its simple model and optimal solutions. But it is difficult to track multi-target from consecutive frames under long-term occlusions. Therefore, an approach combining extended Kalman filter (EKF) and NF algorithm is proposed in this paper, in which EKF can predict the position of the occluded target. Experimental results show that the proposed tracking method based on EKF and NF (EKF-NF), can enhance the stability of tracking multiple moving targets in complex scenarios.