Abstract
Omni-vision system using an omni-mirror is popular to acquire omnidirectional environment information for an autonomous mobile robot. However, the self-localization sometimes becomes unstable because the omni-vision system suffers from occlusion caused by surrounding human and moving objects in a horizontal direction. On the other hand, the visual information of the ceiling above the robot is useful for the self-localization because the information is rarely occluded by moving objects around the robot. Self-localization methods based on the visual information of the ceiling have been proposed so far. We have investigated self-localization methods based on local features of the ceiling vision and found that they need high computational costs and have problems with real-time performance.
In this paper, improvement of reducing processing time based on edge detection of the ceiling image is shown. We investigate the validity of the proposed method from the experiment of autonomous mobile robot in the miniature soccer field.
