基于RGB-D信息的室内占据栅格地图生成方法

To avoid the high cost of lidar, an occupancy grid mapping method for indoor navigation based on RGB-D information is proposed. Based on the color and depth information, a global point cloud map is generated combining with the continuous pose of camera. After pretreatment of downsampling and outlier elimination, the ground and non-ground point cloud are segmented by morphological progressive filter algorithm. The normals are obtained from the re-sampled ground, and the non-ground point cloud is projected into grids. Using bayesian probability method to update the occupied status of grids, and then the occupancy grid map is generated. Experimental results indicate that the method can generate 2D occupancy grid maps at different heights for robotic indoor navigation, remaining the high accuracy and avoiding the high cost.