Laser SLAM method and device based on multi-frame space occupancy rate and medium

The invention discloses a laser SLAM method and device based on a multi-frame space occupancy rate and a medium, and relates to the field of dynamic object recognition in SLAM. The method comprises the steps that a target point cloud image is projected on a horizontal reference plane, and voxel grid division is carried out; carrying out downsampling on all voxel grids based on the distance value corresponding to each voxel grid, and then carrying out cross feature extraction to obtain a spatial feature point set; performing rough matching on the multi-frame continuous laser point clouds based on the spatial feature point set to construct a local map; based on the local map, voting of voxel grids is carried out on the multi-frame continuous laser point cloud to determine dynamic voxel grids and remove corresponding laser points to obtain an optimized point cloud image, and then fine matching is carried out to obtain a preliminary global map; and optimizing the initial global map by adopting a loopback detectio