Robust Object Tracking via Information Theoretic Measures
- PDF / 3,211,203 Bytes
- 15 Pages / 595.26 x 841.82 pts (A4) Page_size
- 84 Downloads / 219 Views
t Object Tracking via Information Theoretic Measures Wei-Ning Wang 1,2 Qi Li 1,2,3 Liang Wang 1,2 1 Center for Research on Intelligent Perception and Computing, National Laboratory of Pattern Recognition,
Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China 2 School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100190, China 3 Artificial Intelligence Research, Chinese Academy of Sciences, Qingdao 266300, China
Abstract: Object tracking is a very important topic in the field of computer vision. Many sophisticated appearance models have been proposed. Among them, the trackers based on holistic appearance information provide a compact notion of the tracked object and thus are robust to appearance variations under a small amount of noise. However, in practice, the tracked objects are often corrupted by complex noises (e.g., partial occlusions, illumination variations) so that the original appearance-based trackers become less effective. This paper presents a correntropy-based robust holistic tracking algorithm to deal with various noises. Then, a half-quadratic algorithm is carefully employed to minimize the correntropy-based objective function. Based on the proposed information theoretic algorithm, we design a simple and effective template update scheme for object tracking. Experimental results on publicly available videos demonstrate that the proposed tracker outperforms other popular tracking algorithms. Keywords: Object tracking, information theoretic measures, correntropy, template update, robust to complex noises.
1 Introduction Object tracking is a very important topic in computer vision. It aims to estimate the spatial state of a moving target in a video sequence[1–9]. With an object track in the first frame identified, the tracking problem is usually formulated as automatically tracking the trajectory of the object over the subsequent frames. It has been widely applied in many real world problems, such as vehicle navigation and video surveillance. However, accurate tracking of general objects under complex scenarios is still difficult due to partial occlusions, illumination variations, abrupt object motions, cluttered backgrounds, etc. Tremendous efforts in object tracking have been made to tackle these problems in recent years[10–15]. Deep learning based methods have shown superior performance over traditional methods on object detection, object segmentation, object recognition[16–20], etc. They have also been widely used for tracking[17,21–23]. Although deep learning based tracking algorithms have achieved big breakthroughs in recent years, they still suffer from heavy computational cost and limited training data. In this paper, we mainly focus on traditional tracking algorithms. There are generally two major categories in tra Research Article Manuscript received February 10, 2020; accepted April 20, 2020 Recommended by Associate Editor Hui Yu © Institute of Automation, Chi
Data Loading...
