An Efficient Compression Method for Collaborative 3D Mapping in Confined Space with Limited Resources

LiDAR is a key sensor for 3D mapping, and as the measurement accuracy of the sensor continues to improve, the order of magnitude and the memory required to acquire 3D point cloud data are getting larger and larger. However, due to the limited computational resources of unmanned platforms for collaborative 3D mapping in confined space, the large memory occupancy of the point cloud data will bring challenges to the subsequent transmission, storage and collaboration. In order to save storage space and improve transmission and processing efficiency, this paper expresses the 3D point cloud based on the voxel octree structure. We use cropping to remove the irrelevant point cloud information first, and further optimize the compression algorithm of the voxel octree by detecting the point cloud resolution and then setting the voxel filter in partitions. Experimental results verify the proposed method compress the point cloud as much as possible while reducing the loss of key information.