Tracking algorithm based on correlation filter have been extensively investigated due to their powerful performance in benchmark datasets and competitions. However, the periodic assumption has contributed boundary effects and the complex scenarios will give rise to model drift, which have an extremely negative effect on both tracking precision and success rate. To mitigate these challenges, a novel multi-model and multi-expert correlation filter (MMCF) approach is proposed in this paper. The key innovation of the proposed method is to employ multiple models and experts for tracking. Multiple models can excavate diverse a large quantity of feature information, and the target information between different models can complement each other. Multiple experts provide several possible predictions, while an evaluation mechanism selects the most reliable prediction utilizing their assessment scores. To further improve the performance, an adaption strategy is utilized to update the multiple models, which can reduce the weight of the bad samples to prevent model drift. Experiments performed on three recent benchmark datasets OTB50, OTB2013, OTB100, TC-128 and UAV123@10fps, have demonstrated the superiority of our approach in comparison to the state-of-the-art trackers. Our MMCF tracker operates a speed of about 58 frames per second (FPS) running on a single central processing unit(CPU). INDEX TERMS Correlation filter; boundary effect, model drift, multiple models, multiple experts.

Non-member Although the correlation filtering tracker for visual target tracking has achieved excellent results in both accuracy and robustness, there are still some problems yet to be solved. Obtaining stable scale estimation using traditional trackers is a challenging problem in visual target tracking, and many trackers fail to handle scale change in complex video sequences. In order to solve the problems of scale change, partial occlusion and geometric deformation for target tracking effectively, a new tracker based on kernel correlation filtering is developed in our study. The tracker obtained with maximum posterior probability method has scale adaptive ability and can deal with scale change to improve the tracking ability. In addition, the tracker further enhances the ability to deal with illumination variation, geometric deformation and occlusion by fusing the adaptive color naming feature and the histogram of oriented gradient feature as well. The VOT-2018 which has 50 video sequences is used as the benchmark data set in this work and the simulation evaluation on this data set have shown that the proposed tracker has achieved stable tracking results in some challenging scenarios and can achieve better tracking performance than other trackers.

No abstract

No abstract

No abstract