Incremental Visual Tracking with l 1 Norm Approximation and Grassmann Update

This chapter proposes a new incremental tracking algorithm based on l1 norm approximation and Grassmann subspace update. Based on previous subspace, the linear approximation on this subspace employs l1 norm, and the problem can be transformed as unconstrained augmented Lagrangian form solved by alternating direction method of multiplier (ADMM) method. The tracking problem is performed in geometrical particle filter on affine group Aff (2). The state model is described by first-order autoregressive (AR) process. And the likelihood is based on the l1 norm approximation error. In order to tackle occlusion issue, the dual vector is introduced into the likelihood. The subspace update can be considered as an optimization problem on Grassmann manifold. The step size along the geodesic is important; an adaptive step-size strategy is given. The experimental results demonstrate that our tracking performance is superior to the other state-of-art trackers under many challenging tracking situations.