Various tracking algorithms based on correlation filters cannot achieve accurate results for the reason of complex factors in the process of tracking,such as background cluster,occlusion,illumination variation,deformation,etc.In view of the shortcoming of traditional correlation filters,an effective tracking framework is proposed by combining multi-angle discriminative correlation filters and keypoints to achieve the motivation of position estimation.Meanwhile,multi-scale discriminative correlation filters are used to estimate the scale of target.Experimental results show that the proposed algorithm can achieve well performance in tracking and outperform state-of-the-art methods by a significant margin.%跟踪过程中的复杂因素,如繁杂背景、遮挡、光照变化、形变等导致了很多相关滤波跟踪算法在跟踪时不能获得很好的精度.针对传统相关滤波的不足,提出了一种有效的跟踪框架,通过联合多角度判别相关滤波与关键点达到目标位置估计的目的,再利用多尺度相关滤波估算目标的尺度.实验结果表明,提出的跟踪算法能够很好地对目标进行跟踪.相比其他的跟踪算法,新算法在精度上有很大的提高.
展开▼
