Assuming that the measurement and process noise covariances are known, the probability hypothesis density (PHD) filter is effective in real-time multi-target tracking; however, noise covariance is often unknown and time-varying for an actual scene. To solve this problem, a strong tracking PHD filter based on Variational Bayes (VB) approximation is proposed in this paper. The measurement noise covariance is described in the linear system by the inverse Wishart (IW) distribution. Then, the fading factor in the strong tracking principle uses the optimal measurement noise covariance at the previous moment to control the state prediction covariance in real-time. The Gaussian IW (GIW) joint distribution adopts the VB approximation to jointly return the measurement noise covariance and the target state covariance. The simulation results show that, compared with the traditional Gaussian mixture PHD (GM-PHD) and the VB-adaptive PHD, the proposed algorithm has higher tracking accuracy and stronger robustness in a more reasonable calculation time.

Multi-resolution feature fusion DCF (Discriminative Correlation Filter) methods have significantly advanced the object tracking performance. However, careless choice and fusion of sample features make the algorithm susceptible to interference, leading to tracking failure. Some trackers embed the re-detection module to remedy tracking failures, yet distinguishing ability and stability of the sample features are scarcely considered when training the detector, resulting in low effectiveness detection. Firstly, this paper proposes a criterion of feature tracking reliability and conduct a novel feature adaptive fusion framework. The feature tracking reliability criterion is proposed to evaluate the robustness and distinguishing ability of the sample features. Secondly, a re-detection module is proposed to further avoid tracking failures and increase the accuracy of target re-detection. The re-detection module consists of multiple SVM detectors trained by different sample features. When the tracking fails, the SVM detector trained by the most reliable sample feature will be activated to recover the target and adjust the target position. Finally, comparison experiments on OTB2015 and UAV123 databases demonstrate the accuracy and robustness of the proposed method.