Multi-Model and Multi-Expert Correlation Filter for High-Speed Tracking

Abstract

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.

Application of Bagging, Boosting and Stacking Ensemble and EasyEnsemble Methods for Landslide Susceptibility Mapping in the Three Gorges Reservoir Area of China

“…A weak learner refers to a learner with generalisation performance slightly better than that for a random guess [ 24 ]. Usually, different weights are assigned according to the classification accuracy, and the samples with low accuracy are given higher weights [ 25 ]. The samples with higher weights are considered by subsequent learners.…”

Multifeature Fusion Tracking Algorithm Based on Self-Associative Memory Learning Mechanism

“…Track-32 ing algorithms can be divided into two categories, depend-33 ing on the appearance model: generative and discriminative. 34 Generative tracking [2] learns online to model directly and 35 then searches for the closest candidate domain to the target 36 with the help of the model to determine the next position of 37 the target. However, the generative class of algorithms does 38 not consider background information regarding the target, 39 which is prone to tracking failure.…”

Research on preview control algorithm of PRS-XY hybrid machine tool based on kinematics calculation

“…Based on the online learning control method, the shortcomings of its slow convergence speed is not conducive to the realization of high-speed and stable motion control [5]. Preview control can be based on a linear control object model, or based on a nonlinear control object model, but requires precise control of the object model [6].…”