MULTI-OBJECT COLOR TRACKING VISION SYSTEM FOR AUTONOMOUS ROBOT

Color provides powerful information for object recognition. But it's difficult for robot vision system to recognize targets quickly and robustly using such information. To solve this problem, this paper describes a color training robot vision system capable of real time tracking multi objects quickly by very low cost commodity hardware and presents a novel but easy color training and object tracking algorithm. The vision system has been successfully integrated with our autonomous robot system and shown its quickness and robustness in dynamic object tracking experiments in non structured environments.