3D Positioning for Mobile Robot Using Omnidirectional Vision

Abstract

This paper introduces a local 3D positioning system using omnidirectional vision for mobile robot navigation in a known indoor/outdoor working field environment. The omnidirectional vision image can provide 360 degree information around the sensor and the direction angle information data with the principle point of camera obtained relatively accurate from the image. The approached system set four red artificial landmarks as a rectangle on the corners of an operating spot. The x and y horizontal location was calculated based on the direction angles of landmarks. The z axis value was calculated via using the relation between sensor height and distance from sensor to landmark. Position experiment was conducted on the field in a 5 m×5 m rectangular area and straight-line driving practical experiments was conducted on a 30 m distance field road. Sixteen test positions were chose on average. Results showed that the distance RMS error between estimated position and test position was about 6.84 cm and maximum error was about 12 cm. The practical experimental results showed that the distance RMS error was about 7.34 cm. In conclusion, the novel positioning system is feasible and the accuracy is good.