This paper presents a new positioning technique, called omnidirectional dynamic vision positioning, for use in navigating a mobile robot. A fisheye lens image system is used as an omnidirectional vision device. A positioning algorithm used with reference beacons determines the position of the robot in terms of a global positioning coordinate system. An accuracy analysis was conducted, and positioning errors were characterized. The experimental results verify the practicability and realizability of the new technique. The advantage of omnidirectional vision for navigation is the simplicity of recording an entire 2 tt radian scene without camera scanning.