Based on by the Dijkstra farthest point sampling (DFPS) algorithm, the initial sampling points of the three dimensional (3D) model cluster are obtained. The correspondence between two models is obtained by the calculation based on the functional mapping theory. The matching between two models is transformed into the multi way correspondence of a 3D model cluster by use of the cycle consistency constraints. The experimental results show that the proposed algorithm can reduce the isometric errors of the correspondence among 3D models to a certain extent. Moreover, this algorithm is not only capable of the correspondence calculation of two models, but also the correspondence calculation of an isometric or approximately isometric 3D model cluster.