Ground Moving Target Tracking for a Patrol Robot Based on Monocular Vision
Qifeng Yang, Fang Xu, Daokui Qu, Yilin Hong, Yan Zhuang · 2017
This paper investigates the problem of monocular-vision based ground moving target tracking for a patrol robot. Considering the patrol robot and the tracking target are both moving, we cannot obtain enough features from image data in real time to ensure the reliability of object detection and tracking. To solve the above problem, we design a ground target tracking system which is based on TLD (Tracking Learning Detection) algorithm. The TLD algorithm integrates tracking, detection and learning modules, which can improve the performance and the robustness of moving target tracking. To show the practicability of this system, a series of experiments were conducted on a real micro patrol robot. Experimental results show the effectiveness of our ground moving target tracking method for a patrol robot.