An improved TLD target tracking algorithm

Target tracking has always been a hot research topic in the field of computer vision. Tracking-Learning-Detection (TLD) is a new algorithm for online learning tracking proposed by Zdenek Kalal. In the algorithm, the computation consuming of detection module is relatively large. To solve this problem and improve the algorithm, we proposed an online learning method to adaptively update the threshold of variance classifier, which can effectively reduce the number of target boxes, improve the real-time performance and tracking accuracy. Experiments have been conducted to compare the performance of the improved TLD with the original TLD. The experimental results show that the improved TLD has better real-time performance and higher accuracy for tracking.

I. Inroduction

Target tracking technology in video is a hot issue in the field of computer vision, which has been widely used in many fields, such as intelligent monitoring, video conferencing, human-computer interaction, traffic control, navigation guidance and so on. It is of important significance for the development of these fields. However, the target tracking technology is faced with many challenges in the practical application, such as illumination change, target deformation, target occluded or disappeared, high-speed moving target, moving camera, complex background and so on. In order to solve these problems, researchers of various countries have proposed a large number of target tracking algorithms. The traditional target tracking algorithm can not deal well with the occlusion, loss and pose change of the target, and it can not continuously guarantee the accuracy and robustness in the process of long time tracking. The current development trend of target tracking is the increasing introduction of the online learning mechanism. Among them, a new TLD framework for online learning tracking is proposed by Zdenek Kalal [1] which needs less priori information for long term online tracking. And it also has a good performance in the case of the target occluded or disappeared, as well as the appearance change of the target. Therefore, it has received much attention, and many researchers have begun to study and analyze it [2]–[5]. Aiming at the problem of large computation consuming in detection module of TLD algorithm, we proposed an online learning method to adaptively update the threshold of variance classifier. The improved TLD not only improves the real-time performance, but also improves the tracking accuracy of algorithm.