Abstract

Appearance models play an important role in visual tracking. Effective modeling of the appearance of tracked objects is still a challenging problem because of object appearance changes caused by factors, such as partial occlusion, illumination variation and deformation, etc. In this paper, we propose a tracking method based on the patch descriptor and the structural local sparse representation. In our method, the object is firstly divided into multiple non-overlapped patches, and the patch sparse coefficients are obtained by structural local sparse representation. Secondly, each patch is further decomposed into several sub-patches. The patch descriptors are defined as the proportion of sub-patches, of which the reconstruction error is less than the given threshold. Finally, the appearance of an object is modeled by the patch descriptors and the patch sparse coefficients. Furthermore, in order to adapt to appearance changes of an object and alleviate the model drift, an outlier-aware template update scheme is introduced. Experimental results on a large benchmark dataset demonstrate the effectiveness of the proposed method.

Highlights

  • Visual tracking is a hot topic in the field of computer vision and has a wide range of applications, such as vision-based control [1], unmanned aerial vehicles [2], intelligent transportation [3], etc. some significant progress has been made in recent years, it still remains a challenging problem due to numerous appearance changes caused by factors, such as illumination changes, occlusion, scale variations, shape deformation, etc

  • Based on structural local sparse representation in [9], we design the patch descriptor to reflect the degree, to which each patch is contaminated with noise caused by appearance changes

  • In order to reduce the risk of model drift, they construct a refiner model based on an online support vector machine (SVM) detector to update an incorrect prediction to a reliable position in case of low reliability of the current tracking result

Read more

Summary

Introduction

Visual tracking is a hot topic in the field of computer vision and has a wide range of applications, such as vision-based control [1], unmanned aerial vehicles [2], intelligent transportation [3], etc. Formulated tracking as a sparse approximation problem This method employs holistic representation schemes and does not perform well when target objects are heavily occluded. In [9], a tracking method based on the structural local sparse appearance model was proposed. This model exploits both partial information and spatial information, and handles occlusion by pooling across the local. Based on structural local sparse representation in [9], we design the patch descriptor to reflect the degree, to which each patch is contaminated with noise caused by appearance changes.

Related Work
Patch-Based Tracking Methods
Strategies for Alleviating Model Drift
Target Region Division
Structural Local Sparse Representation
Object Tracking
Update Scheme
Experiment
Overall Performance
Attribute-Based Analysis
Precision
Evaluation of Template Update Strategy
Typical Results Analysis
Conclusions
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call