Due to the continuous development of 3D scanning technology, the point cloud data is becoming more and more redundant. Storing or handling these data will consume a lot of time and computer resources. In this paper, a method for cloud data reduction based on 3D Grids is proposed. It includes searching k-nearest neighbors for constructing data topology, calculating tangent plane normal, sampling points using uniform 3D grids and selecting points that satisfy the requirements. The method retains more features than traditional methods, and it avoid the estimating of curvature.