A new local monocular visual map approach suitable for robot navigation and scenario understanding

This work introduces a new methodology that can be used as a local map for the navigation of robots. Moreover, environment structure can be inferred. The objectives are attained by using monocular techniques and odometry data, within a local occupancy grid framework. The monocular field of view is constrained to the vicinity of the mobile robot. In this way, a simplified depth from focus method, which uses only one frame, is implemented; hence, scenario information can be achieved when homogeneous radiance background constraint is accomplished. Further structure analysis can be developed by computing optical flow through time-integration of acquired frames within a tessellated probabilistic representation consisting of a local occupancy grid framework. The camera pose knowledge and odometer data are used to correlate the different overlapping image zones. Moreover, time-integration of the monocular frames allows a larger environment description suitable for WMR local path planning.