In point clouds reduction, the uniform grid method can achieve rapid reduction of point clouds data, but is not sensitive to changes in object shape and curvature. Different adaptive grid sizes are proposed based on object morphology and point clouds curvature, and redundant point clouds data is deleted by median filtering. It not only inherits the fast and effective advantages of the uniform grid method, but also preserves object detail information. Compared to the uniform grid method, the detailed features are enhanced and the reduction efficiency is improved by 12.8%.