Lasso约束下融和光流信息的DCF目标跟踪算法

DCF Visual Object Tracking Algorithm with Lasso Constraints and Fusion Optical Flow

视频目标跟踪是计算机视觉的基础问题之一。近来由于discriminative correlation filter(DCF)跟踪器的高效性和鲁棒性,出现了许多基于DCF的目标跟踪算法。为了克服DCF跟踪器对运动模糊目标的不适应性,本文提出了一种利用Lasso约束并融入光流信息的目标跟踪算法。首先在跟踪器抽取特征通道块中融入光流特征。然后在通道块之后进行多特征融合。其次利用Lasso约束DCF跟踪器的目标函数。考虑到所约束的目标函数在定义域上不连续和目标跟踪的优化效率。最后,采用块坐标下降算法来优化所约束的目标函数。实验结果表明,与基于DCF视觉跟踪算法相比,所提出的算法可以有效的处理运动模糊目标,实现复杂环境下鲁棒的视觉目标跟踪。

Video object tracking is a fundamental problem in computer vision.Recently,due to the high efficiency and robustness of the discriminative correlation filter(DCF),many DCF-based object tracking algorithms have emerged.In order to overcome the incompatibility of DCF tracker to motion blur object,this paper proposes a new algorithm that exploits Lasso constraints and integrates optical flow.First,the optical flow is integrated into the proposed tracker.Secondly,the objective function of DCF was constrained by Lasso.Consider the optimization efficiency of the constrained objective function,the block coordinate descent algorithm is used to optimize the constrained objective function.Compared with the DCF-based algorithms,the experimental results show that our algorithm can effectively track motion blur object and achieve robust tracking in complex environments.