This paper is mainly aimed at the over-segmentation or under-segmentation phenomenon of the ground point cloud in the moving scene of the unmanned platform to identify the target. This paper proposes an optimized rasterization method to remove the ground point clouds appropriately. First, the partitioned data by quadrant is processed based on the asymmetric distribution of front-to-back and left-to-right point cloud data. Then, the ground estimation of the lowest point is carried out by drilling into the laser layer data, and the connection between the estimated points and the ground detection point cloud data is segmented using the road reflection intensity correction. Next, the ground point cloud detection filtering is optimized using the raster elevation difference. Finally, an obstacle continuum hypothesis model is used to improve the under-segmentation phenomenon that occurs inside the raster. The overall ground point cloud detection filtering effect, the algorithm has a certain degree of universality and ideal ground detection filtering effect within the road, the overall achieve the desired goal.