The reduction and storage of enormous point cloud data is a crucial link in reverse model reconstruction. Considering the features of point cloud data by single station laser scanning, a new method — grid sector method was put forward for its reduction and storage. Point cloud data could be filtered and stored only by traversal. This method was realized on VC++ 6.0. Multi-station scanning of point cloud registration and stitching would be more quickly and efficiently, if the site goes through the fan in a single grid after treatment. Based on the contrast with traditional compressing methods, this paper analyzed its characteristics and proved its applicability in battlefield terrain digitization.