Abstract: Proposing a new point cloud data simplification method.At first,calculate the normal vectors of all the points and the average normal vector of all points in the bounding box,whether or not to subdivide the cube using octree grid division algorithm depends on the angle between the two vectors.Secondly,the point cloud is fitted using quadratic parametric curve method and principal curvatures of all points are estimated,and Hausdorff distance of principal curvatures decides to get and whether to keep the feature points.Experiments show that simplication results are obvious,can delete the redundant points in the meantime reserve the original cloud geometry features,and increases the three dimensional reconstruction efficient.