No abstract

No abstract

For the problem of robust visual tracking in various complex tracking scenarios, a multi-cue correlation particle filter (CPF for short) visual tracker supervised by population convergence is proposed. By combining the advantages of particle filter and correlation filter, the CPF tracker gains better robustness, computational efficiency and stability for visual tracking. Meanwhile, to solve the problem of sample diversity in traditional CPF tracker, a genetic operating algorithm supervised by population convergence is proposed and introduced to the resampling process of CPF. Then considering that a single kind of feature weakens the tracking efficiency and robustness of our tracker, we propose to combine different types of features including Harris feature, HOG feature and SIFT feature based on fuzzy control theory to form a multi-cue CPF tracker (SPC-MCCPF for short). Multiple experiments on the OTB2015 and VOT2018 datasets prove that our tracker is quite effective in dealing with various challenging tracking problems. INDEX TERMS Computer vision, target tracking, feature extraction, correlation particle filter. I. INTRODUCTION Visual target tracking has a wide range of applications in hotspots such as traffic monitoring, automatic driving, human-machine interaction, violence identification, suspect inspection and other fields [1]-[3]. Although great progress has been made in recent years, there still remain various challenging problems such as low resolution, deformation and occlusion [4]. This study therefore focuses on dealing with these challenges and develops a robust and efficient tracking method. Recently, particle filter has been applied in the field of visual tracking because it is suitable for solving nonlinear problems like nonlinear target motion and is flexible in combination with various kinds of target representations. Ref. Concha et al. [5] presented an adequate performance analysis of the particle filtering (PF for short) algorithm for a computationally intensive 3D multi-view visual tracking problem. Zhang et al. [6] proposed a hybrid positioning method based on particle swarm optimization algorithm and particle filter. This method is mainly used for positioning tasks with a priori environment map. This research obtains a robust positioning method based on PF, which can work in symmetrical environments. Wang et al. [7] developed a new tracking method, The associate editor coordinating the review of this manuscript and approving it for publication was Juan A. Lara .

For problems related to the robust tracking of visual objects in various challenging tracking conditions, a robust visual tracking method based on multilayer convolutional features and correlation filtering is proposed. To solve the problems of mean deviation and insufficient discrimination ability in traditional convolutional neural networks (CNN), this study proposes randomized parametric rectified linear units (RPReLU) as the activation function. Meanwhile, the zero-setting operation of weights in the traditional dropout process occurs randomly and fails to discriminate the features with different weights, which leads to a low learning efficiency. Therefore, this study proposes an improved dropout method based on a support vector machine (SVM), which provides a selective dropout rate to increase the manual orientation and improve the learning efficiency of the dropout process. In addition, traditional CNN trackers only employ the output of the last layer, which can effectively capture semantic features but not spatial features. To solve this problem, we propose to use the rich features of the multiple convolution layers of CaffeNet as the target representation. Furthermore, we propose an improved correlation filter to further improve the tracking performance and improve the tracker's capability of dealing with scale changes, which effectively solves the problem of adaptive estimating of target size. The extensive experimental evaluations have been carried out through the OTB2015, VOT2016 and VOT2018 datasets, proving that the proposed method is very effective in dealing with a variety of challenging factors.INDEX TERMS Convolutional neutral network, correlation filter, target tracking, computer vision technology.

No abstract