Attention region detection and tracking for observing a person manipulating day-to-day objects
Kotaro Nagahama, Tomohiro Nishino, Kimitoshi Yamazaki, Kei Okada, Masayuki Inaba · International Journal of Mechatronics and Automation · 2013
This paper presents a method for robots tracking multiple important regions with proper resolution while the robot is observing a person in daily environments. The system consists of a visual function to track end points and a composite vision system. The visual function uses 'continuous disparity filtering' that is an expansion of zero disparity iltering and other evaluation functions to track a person's head, hands and tips of tools. Besides, we developed a composite vision head with wide–angle cameras and zoom–enabled stereo cameras. We tested the visual function and the vision system by experiments to track a human head and hands, and to change the attention position to be captured with high resolution. These results showed the effectiveness of our approach.