基于稀疏稠密结构表示与在线鲁棒字典学习的视觉跟踪

Visual Tracking Based on Sparse Dense Structure Representation and Online Robust Dictionary Learning

L1跟踪对适度的遮挡具有鲁棒性,但是存在速度慢和易产生模型漂移的不足。为了解决上述两个问题,该文首先提出一种基于稀疏稠密结构的鲁棒表示模型。该模型对目标模板系数和小模板系数分别进行L2范数和L1范数正则化增强了对离群模板的鲁棒性。为了提高目标跟踪速度,基于块坐标优化原理,用岭回归和软阈值操作建立了该模型的快速算法。其次,为降低模型漂移的发生,该文提出一种在线鲁棒的字典学习算法用于模板更新。在粒子滤波框架下,用该表示模型和字典学习算法实现了鲁棒快速的跟踪方法。在多个具有挑战性的图像序列上的实验结果表明:与现有跟踪方法相比,所提跟踪方法具有较优的跟踪性能。

The L1trackers are robust to moderate occlusion. However, the L1trackers are very computationally expensive and prone to model drift. To deal with these problems, firstly, a robust representation model is proposed based on sparse dense structure. The tracking robustness is improved by adding an L2norm regularization on the coefficients associated with the target templates and L1norm regularization on the coefficients associated with the trivial templates. To accelerate object tracking, a block coordinate optimization theory based fast numerical algorithm for the proposed representation model is designed via the ridge regression and the soft shrinkage operator. Secondly, to avoid model drift, an online robust dictionary learning algorithm is proposed for template update. Robust fast visual tracker is achieved via the proposed representation model and dictionary learning algorithm in particle filter framework. The experimental results on several challenging image sequences show that the proposed method has better performance than the state-of-the-art tracker.