Landmark design and real-time recognition based on fisheye image for robot navigation

Robot navigation has been a hot topic in the area of mobile robot. A lot of research has been undergone ranging from noble algorithms to applications. Fisheye image has wide angle View and it is easy to obtain the position of the object and keep it in vision within large scale, so it is an important research domain in robot navigation. However it has a disadvantage that the image obtained from fisheye has severe distortions which improve the difficulties in object detection. In this paper we design a circle landmark and combine the feature of the color and edge of the landmark to identify and track. To improve the robustness we still design a feature update model in the tracking process. Experiment shows that this method can recognize and track the landmark robustly and implement robot navigation in room environment effectively.