Online Learning of Discriminative Correlation Filter Bank for Visual Tracking

Jian Wei,Feng Liu

doi:10.3390/info9030061

Abstract

Accurate visual tracking is a challenging research topic in the field of computer vision. The challenge emanates from various issues, such as target deformation, background clutter, scale variations, and occlusion. In this setting, discriminative correlation filter (DCF)-based trackers have demonstrated excellent performance in terms of speed. However, existing correlation filter-based trackers cannot handle major changes in appearance due to severe occlusions, which eventually result in the development of a bounding box for target drift tracking. In this study, we use a set of DCFs called discriminative correlation filter bank (DCFB) for visual tracking to address the key causes of object occlusion and drift in a tracking-by-detection framework. In this work, we treat thxe current location of the target frame as the center, extract several samples around the target, and perform online learning of DCFB. The sliding window then extracts numerous samples within a large radius of the area where the object in the next frame is previously located. These samples are used for the DCFB to perform correlation operation in the Fourier domain to estimate the location of the new object; the coordinates of the largest correlation scores indicate the position of the new target. The DCFB is updated according to the location of the new target. Experimental results on the quantitative and qualitative evaluations on the challenging benchmark sequences show that the proposed framework improves tracking performance compared with several state-of-the-art trackers.

Highlights

In recent years, numerous visual object tracking (VOT) algorithms have been developed to overcome the limitations of VOT; these methods provide technical support for practical applications; this topic is clearly a popular research direction in the field of computer vision, and it has important applications in various areas, such as intelligent surveillance systems, human–computer interaction, autonomous driving, unmanned aerial vehicle (UAV) monitoring, video indexing, and intelligent traffic monitoring [1]
The existing algorithm based on discriminative correlation filters (DCFs) causes the bounding box to deviate from the target when encountered with partial or full occlusion in complex scenarios
We address this limitation by learning the discriminative correlation filter bank (DCFB) through the extracted samples of the current frame and accurately estimating the location of the target in the frame

Summary

Introduction

Numerous visual object tracking (VOT) algorithms have been developed to overcome the limitations of VOT; these methods provide technical support for practical applications; this topic is clearly a popular research direction in the field of computer vision, and it has important applications in various areas, such as intelligent surveillance systems, human–computer interaction, autonomous driving, unmanned aerial vehicle (UAV) monitoring, video indexing, and intelligent traffic monitoring [1]. A major breakthrough has been made in theoretical research, the design of a robust tracking system encounters numerous difficulties in practical complex scenarios, such as illumination variation, scale variation, occlusion, deformation, motion blur, rapid motion, in-plane rotation, out-of-plane rotation, out-of-view condition, background clutter, and low resolution. The existing algorithm based on discriminative correlation filters (DCFs) causes the bounding box to deviate from the target when encountered with partial or full occlusion in complex scenarios. We focus on the challenge posed by occlusions to object tracking. We address this limitation by learning the discriminative correlation filter bank (DCFB) through the extracted samples of the current frame and accurately estimating the location of the target in the frame. Generative trackers [2,3,4]

Objectives

Methods

Results

Conclusion