摘要： 为提高目标跟踪算法在复杂条件下的鲁棒性和准确性,在粒子滤波框架下,提出一种基于稀疏表示的2范数最小化时间连续目标跟踪算法。使用模板字典线性重构候选目标,利用2范数对重构系数进行稀疏约束,构建2范数约束的目标方程。考虑到目标状态帧间残差的连续性,将残差连续约束项嵌入目标方程,通过求偏导数的方法求解目标方程。采用稀疏表示与增量学习结合的方法对模板字典进行更新,实现目标的精确跟踪。实验结果表明,与现有目标跟踪算法相比,该算法具有较强的跟踪鲁棒性及抗干扰能力。

关键词: 目标跟踪, 稀疏表示, 粒子滤波, 模板字典, 残差, 连续性

Abstract: For the purpose of improving the robustness and accuary of target tracking algorithms in complicated conditions,under the particle filtering framework,this paper proposes a continuous time target tracking algorithm of 2-norm minimizaiton based on sparse representation.Firstly,candidate targets are linearly reconstructed by a template dictionary,and the coefficients are sparsely constrained by 2-norm.As a result,a 2-norm constraint objective function is built.Secondly,the proposed tracker takes the temporal consistency of target state inter-frame residual into account,and embeds a residual consistency constrain term into the objective function.The solution of the objective function is designed by taking the partial derivative.Thirdly,a method which combines sparse representation and principal component analysis is used for updating the dictionary to realize accurate targer tracking.Experimental results show that the proposed tracking algorithm has better tracking robustness and anti-interference ability than existing tracking algorithms.

Key words: target tracking, sparse representation, particle filtering, template dictionary, residual, continuity

中图分类号: 

• TP319.41