An aerodynamic drag reduction optimization design study of high-speed train head is carried out based on the three-dimensional parametric approach of local shape function, improved ant colony algorithm and improved Kriging surrogate model. To avoid repeated generation of ten millions of meshes in the case of large deformation with complex geometry and improve the optimization efficiency of high-speed train head, we introduce mesh deformation techniques of the reduced control points based on radial basis functions (RBF). The optimization results show that: RBF mesh deformation method could largely shorten the time-consuming of mesh deformation without reducing the quality of meshes, and can be used for aerodynamic optimization design of complex geometry. Under the design space given in this article, the six key design parameters that control the nose shape have effects on the aerodynamic drag of the train with a kind of monotonically increasing relationship. After optimization under the constraints, the total aerodynamic drag of the simplify shape is reduced by 5.68%. The aerodynamic drag of leading and trailing cars reduced a lot, while the aerodynamic drag of middle car changes little.