BIRD'S EYE VIEW REMAPPING AND PATH FOLLOWING BASED ON OMNIDIRECTIONAL VISION

Abstract | We propose a general method to remap omnidirectional images to bird's eye view images. The described method can be used with dierent omnidirectional systems because it is based on a general projection model. Besides having a large eld of view, the remapped images are orthographic projections of the ground plane. These properties are explored to develop a mobile robot path following task. The control strategy corrects robot's position and orientation while the robot moves inside a lane, limited by one or two white lines on the ground oor. The main contributions are the general bird's eye view remapping method and a simple path following strategy that explores bird's eye view features. The simplicity and the good performance of the controller illustrates the remapping advantages and its applicability