Real-time visual tracking using compressive sensing

An Efficient Misalignment Method for Visual Tracking Based on Sparse Representation

“…Li et al [8] constructed an object representation model based on the RIP (restricted isometry property) of compression perception theory, and utilized Orthogonal Matching Pursuit to tackle L1 minimization problem, Mei et al [9] put forward BPR-L1 tracker which selected particles by minimum error bound, to a large extent reduced the number of particles during the 1 minimization. Bai [14] modeled the appearance of an object as a sparse linear combination of structured union, to solve the sparse representation issue he adopted the BOMP (Block Orthogonal Matching Pursuit) algorithm.…”

“…However, the method used in [4] for solving 1 minimization cost too much time to being real time. Then L1 tracker with minimum error bound was proposed by [8], particles were selected by the minimum error bound and accelerate the resampling procedure. Moreover, instead of updating template each frame, occlusion detection was added up before template update.…”

Consistent multi-layer subtask tracker via hyper-graph regularization

“…Liu et al [33] integrate the dynamic group sparsity into the tracking problem and high dimensional image features are used to improve tracking robustness. In Li et al [34] , dimensionality reduction and a customized orthogonal matching pursuit algorithm are adopted to accelerate the L1 tracker. Zhong et al [5] propose a robust object tracking algorithm using a collaborative model that combines a sparsity-based discriminative classifier (SDC) and a sparsity-based generative model (SGM), but it adopts the naive model updating strategy and similar metric measure, this will affect the performance of the tracker.…”

Discriminative multi-task objects tracking with active feature selection and drift correction

“…Although recent efforts have been made to speed up this tracking paradigm [27,34], these methods assume that sparse representations of particles are independent and ignore their relationships, which can help and improve the tracking performance.…”

“…This representation has been shown to be robust against partial occlusions, which improves the tracking performance. Recently, based on the milestone work, there are several methods have been proposed to improve the L1 tracker in terms of both speed and accuracy [27][28][29][30][31][32][33][34][35][36], such as using accelerated proximal gradient algorithm [29], replacing raw pixel templates with orthogonal basis vectors [32,33], modeling the similarity between different candidates [37], to name a few. Despite of demonstrated success, the above mentioned L1 trackers have the following shortcomings.…”