An Adaptive Resolution OctoMap Method based on Point Cloud Density Analysis

Abstract

OctoMap is an effective 3D occupancy grid mapping approach for indoor robot positioning and navigation. But the fixed resolution setting of OctoMap restricts the environment expression ability and occupies large storage space. To overcome the restriction, an adaptive resolution OctoMap method based on point cloud density analysis is proposed. Firstly, the minimum circumscribed rectangle of the point cloud through PCA analysis. The minimum circumscribed rectangle of the point cloud is equally divided into N point cloud bounding boxes. A density analysis method is proposed to evaluate the environmental complexity of each point cloud bounding box. Finally, the resolution of the OctoMap is adaptively set based on the density analysis of the point cloud bounding boxes. Numerical experiments indicate that the proposed mapping method can self-adaptively adjust the resolution of Octotree map to express environment with various complexity, which effectively solves the contradiction between the environment expression ability and the mapping efficiency.