Planar-Based Visual Positioning for a Mobile Robot with Monocular Vision

Positioning is a fundamental component in many applications such as mobile robot. In this field, vision-based positioning becomes a popular way. This paper proposes an approach of positioning in the planar scene through monocular vision considering its high accuracy and low cost. The accuracy of monocular vision methods depends on the calibration when locating by use of the image of a single camera. The I-C transformation model substitutes arc for edge to reduce approximation error. Furthermore, the C-W transformation model is designed aiming at the absolute coordinates in the planar scene frame with two points as a prior knowledge. A physical testbed is constructed and experimental results reveal that 82% of the error in distances of the I-C transformation model is less than 1 cm and the average is 0.6591 cm. And a theoretical analysis is executed for the worst case performance of the C-W transformation model.