Adaptive Context-Aware and Structural Correlation Filter for Visual Tracking

Abstract

Featured Application: There are many application scenarios for tracking. For example, tracking can be used to assist driving to improve driving safety. Moreover, tracking can assist the police to search for suspicious vehicles in massive videos.Abstract: Accurate visual tracking is a challenging issue in computer vision. Correlation filter (CF) based methods are sought in visual tracking based on their efficiency and high performance. Nonetheless, traditional CF-based trackers have insufficient context information, and easily drift in scenes of fast motion or background clutter. Moreover, CF-based trackers are sensitive to partial occlusion, which may reduce their overall performance and even lead to failure in tracking challenge. In this paper, we presented an adaptive context-aware (CA) and structural correlation filter for tracking. Firstly, we propose a novel context selecting strategy to obtain negative samples. Secondly, to gain robustness against partial occlusion, we construct a structural correlation filter by learning both the holistic and local models. Finally, we introduce an adaptive updating scheme by using a fluctuation parameter. Extensive comprehensive experiments on object tracking benchmark (OTB)-100 datasets demonstrate that our proposed tracker performs favorably against several state-of-the-art trackers.

rStaple: A Robust Complementary Learning Method for Real-Time Object Tracking

“…Most state-of-the-art tracking algorithms adopt the tracking-by-detection method, which is a form of discriminative model. A discriminative model method [18][19][20][21][22][23][24][25][26][27][28][29][30] treats object tracking as an object detection task in each frame wherein image features are collected to train a classifier. Specifically, the target area and background are regarded as positive samples and negative samples, respectively.…”

“…Recently, correlation filters [18][19][20][21][22][23][24][25][26][27] and deep learning-based trackers [28][29][30][31][32][33][34][35] have become two active research topics of discriminative model methods with outstanding performance. Owing to the attractive property of extracting deep features, trackers based on deep learning have demonstrated great potential for the object tracking task.…”

“…For preventing severe tracking failures, a learned detector is used to recover the target position. Recently, Zhou et al [25] applied adaptive context awareness and trained a structural correlation filter for tracking, to make the tracker learn more context information. Shin et al [26] proposed a failure detection and re-tracking algorithm based on KCF performed in real-time.…”

“…The core of correlation filter-based trackers [18][19][20][21][22][23][24][25][26][27] is that they adopt the circulant matrix to construct training samples. Since the correlation operation corresponds to the dot product operation in the Fourier domain, the circulant matrix can be diagonalized by Fourier transform.…”