A Registration Framework for Air-ground Point Cloud Maps

Due to different perspectives, there are significant discrepancies between air-ground point cloud maps in density and blind area, which is a huge challenge for map registration of air-ground collaborative unmanned system. To solve these problems, we design a point cloud map registration framework based on feature error between maps. Firstly, we extract and fuse the features of point cloud maps from aerial and ground. Then the feature error between fusion features and air features is calculated and the optimal registration from ground-map to air-map is estimated by minimizing feature error. Finally, experimental report the accuracy under the initial rotation error of 80 degrees and random translation.