GardenMap: Static point cloud mapping for Garden environment

Building a static environment map is critical in the field of robotics for agriculture. Previous works of static environment mapping algorithm are focus on urban environment, resulting the lack of such algorithm in the garden environment. This paper aims to building static point cloud maps in the garden environment for robot localization and navigation. Towards this goal, we propose a novel static point cloud mapping algorithm for the garden environment, named GardenMap. Specifically, we first processed the point cloud of each frame based on the semantic information, and then we constructed a static point cloud map from the point cloud sequence in the multi-time dimension and multi-resolution fashion.Extensive experiments on two large-scale SLAM datasets validate the effectiveness of our GardenMap, the preservation rate (PR) and rejection rate (RR) of our proposed method outperform state-of-the-art (SOTA) static point cloud mapping algorithm by a large margin.