Abstract: For the discrete and irregular data points，a new method of3D modeling and segmentation was proposed.By using the superquadric parametric models as initial seeds，the best initial seeding state was obtained based on the （μ＋λ） evolutionary strategy，then the nearest neighboring approach was utilized for efficient merging and segmentation of those best initial seeds.Experimental results show that the proposed method can not only achieve the effective modeling and segmentation，but also obtain the topologic relation among the object parts，which is useful for recognition of 3D objects.