基于快速尺度估计的核相关滤波 目标跟踪

为了解决传统基于核相关滤波器（KCF）的跟踪算法难以有效处理目标尺度变化的难题，本文提出了一种新的融合快速准确估计目标尺度变化的核相关滤波跟踪算法。该方法首先利用目标尺度变化的连续性对目标的尺寸变化进行粗略估计，得到目标尺度变化的粗略值；然后进一步对目标尺度的更多可能变化进行精确搜索，提升目标尺度估计的准确性。在公开的复杂场景视频进行测试，比较了本文方法和原始KCF算法的实验效果，并且将本文算法和经典跟踪算法进行了比较，实验结果表明本文提出的目标跟踪算法更准确鲁棒。

Fast Scale Estimation for Object Tracking based on Kernelized Correlation Filter

To address difficulties of traditional kernelized correlation filter (KCF) based tracking algorithm which cannot handle scale-variant object, a new KCF based tracking algorithm which integrates fast scale estimation is proposed in this paper. First the proposed method obtains the coarse scale estimation of the object based on consistency of scale variation of the object during tracking. Then accurate scale estimation is obtained by searching more precise probable scale variations, which can promote the accuracy of scale estimation of the object. We test our method in public videos with complex scenes. We compare results obtained by our method and traditional KCF based tracking algorithm. And we also compare our method with other classic tracking algorithms. Experimental results show that the proposed object tracking algorithm is more robust and accurate.