Obstacle detection in stereo bird's eye view images

Navigation is an essential function for a mobile robot to make it navigate in its environment. The navigation based on bird's eye view images is a newly proposed indoor visual navigation framework, which can help to simplify the navigation problems, such as self-localization, map-building, path planning and other competences. In the paper, an obstacle detection method is proposed, which is applied in the stereo bird's view image based indoor visual navigation. Firstly, the mathematic principle of obstacle detection is derived by using four camera assumptions and imaging geometric relationship. Then the algorithm implementation was explained. Finally, practical experiments were implemented and the results proved that the proposed method can work well.