A self-adaptive alternation of video tracking modes governed by detection of online Kalman performance optimality

Aiming to solve the compromised tracking quality problems arising from solely using the algorithm of Kalman filtering in the circumstances of dynamic modeling mismatch, an approach of tracking mode switch is proposed with the particle filtering used as a stand-by alternative. Two parameters are defined and employed to online supervise the Kalman filter´s performance optimality and self-adaptively activate the tracking mode switch as preset switching criteria are satisfied. The proposed method is put to test on the given lab-synthesized video sequence, suggesting that the tracking quality is significantly bettered as a result of online implementation of tracking mode alternation between Kalman filtering and particle filtering algorithms.