Covariance-based online validation of video tracking

A novel approach is proposed for online evaluation of video tracking without ground-truth data. The temporal evolution of the covariance features is exploited to detect the stability of the tracker output over time. A model validation strategy performs such detection without learning the failure cases of the tracker under evaluation. Then, the tracker performance is estimated by a finite state machine determining whether the tracker is on-target (successful) or not (unsuccessful). The experimental results over a heterogeneous dataset show that the proposed approach outperforms related state-of-the-art approaches in terms of performance and computational cost.