HyperOmni vision: Visual navigation with an omnidirectional image sensor

Abstract

For robot navigation, we need sensor information as to the condition of the environment around the robot. In this paper, we describe an omnidirectional vision system (HyperOmni Vision) using a hyperboloidal mirror and describe its distinctive features, together with a visual navigation method using HyperOmni Vision. In existing omnidirectional vision systems, since the optical system does not use center projection, an independent method of acquiring the visual information is required. However, since the omnidirectional vision system using a hyperboloidal mirror has an omnidirectional visual field and the optical characteristic of center projections, its images can be transformed into images taken directly with an ordinary camera. For this reason, it can produce images suitable for processing, and not only independent techniques, but also conventional image processing technology can be utilized. We will report here the configuration, distinctive features, and characteristics of the optical system of HyperOmni Vision and describe a mobile robot using HyperOmni Vision and its techniques of movement estimation and obstacle detection by template matching. © 1997 Scripta Technica, Inc. Syst Comp Jpn, 28(4): 36–47, 1997