针对传统相关滤波跟踪算法在遮挡、出视野、形变、背景杂乱等挑战场景中容易出现跟踪失败的问题, 本文提出快速尺度支持相关滤波跟踪算法。首先, 利用循环样本构造支持相关滤波器, 并将跟踪问题视为支持相关滤波器的学习问题; 然后, 利用离散傅里叶变换与迭代优化策略解决了支持相关滤波器的学习问题, 极大地降低了算法复杂度; 同时, 利用对数极坐标变换将目标尺度估计问题转换为对数极坐标下的位移变化, 实现了目标尺度的自适应; 最后, 采用自适应模板更新策略, 解决了遮挡情况的模板漂移问题。利用标准数据集测试对本文算法性能进行评估, 结果表明: 所提算法精确度为0.964, 成功率为0.892, 均优于传统的相关滤波跟踪算法, 可以较好地解决形变、遮挡、出视野、背景杂乱等情况下的目标跟踪问题。

The traditional correlation filter tracking algorithm is easy to track failure in the occlusion, out of view, deformation, background clutter, etc. Aiming at this problem, a fast-scale support correlation filter tracking algorithm was proposed in this paper. Firstly, the support correlation filter was constructed by loop sample, and the tracking problem was considered as a learning problem of the support correlation filter. Then, the discrete Fourier transform and the iterative optimization strategy were used to solve the learning problem of the support correlation filter, it greatly reduced the complexity of the algorithm. At the same time, the problem of scale estimation was transformed into displacement variation in polar coordinates by log polar transformation, and the adaptive scale of target was realized. Finally, an adaptive template updating strategy was adopted to solve the template drift problem in occlusion. The performance of the proposed algorithm was evaluated by using standard database. The results show that the accuracy of the proposed algorithm is 0.964 and the success rate is 0892, which are superior to the traditional correlation filtering tracking algorithm. It can better solve the problem of target tracking under the conditions of deformation, occlusion, out of view, and background clutter.