Autonomous Mobile Robot Navigation based on Wide Angle Vision Sensor with Fovea

摘要

In this paper it is presented that the author has examined Wide Angle Foveated Vision Sensor (WAFVS) with a special optical system designed by mimicking a human eye and has advanced it to a parallel processing system by plural computers with multi-task operation system (OS) under LAN network to make the most of its characteristics. WAFVS system carries out parallel image processing flexibly by plural computers with multi-task OS, based on timely task distributing. This system is applied to a navigation system for an autonomous mobile robot. As an example of the execution, experiments of navigation are carried out to obtain an image with sufficient resolution for face recognition, while a camera gazes at and tracks a human face by a proposed method characterized by WAFVS. In this method, central vision using color information, and peripheral vision using object motion information from gray scale images are combined cooperatively.