Global map building based on occupancy grids detected from dense stereo in urban environments

A method for global map building from occupancy grids is presented in this paper. Occupancy grids provide a low-level representation of the environment, suitable for autonomous navigation tasks, in urban driving scenarios. The occupancy grids used in our approach are computed with a method that outputs an occupancy grid with three distinct cell types: road, traffic isles and obstacles. First, we perform a temporal filtering of the false traffic isles present in the grids. Obstacle cells are separated into static (probably infrastructure) and dynamic. An enhanced occupancy grid is built, containing road, traffic isle, static obstacle and dynamic obstacle cells. The global map is obtained by integrating the enhanced occupancy grid along several successive frames. It can be used in various ways, such as alignment with external maps, or for terrain mapping.