A Visual Navigation Method of Mobile Robot Using a Sketched Semantic Map

A new visual navigation method for a mobile robot is proposed in this paper. Its originality lies in integrating a sketched map with a semantic map together for the robot's navigation and in using unified tags to help recognize landmarks. In this sketched semantic map, the outline and semantic information of key referenced objects are used to represent themselves and a rough route for the robot's navigation is also sketched. Over the course of robot navigation along the route, and in order to easily recognize the referenced objects from the complex background, a kind of unified label is designed and pasted on the potential referenced objects in advance. A recognition method based on the Pseudo-Zernike Moment and the Normalized Moment of Inertia is used to compute the matching degree between the real-time image of the referenced object and its similar outline in a database. In addition, the odometer information is also fused so as to roughly localize the robot. Finally, through a series of experiments, the advantage and efficiency of the new navigation method with real-time dynamic obstacle avoidance are testified with the help of the imprecise real map and route