Abstract: The problem of selecting a viewpoint of a 3D object which is the most informative and corresponds to the subjective tendency of most people is known as the best view problem.We propose a method for selecting the best view of 3Dshapes that have been pre-segmented based on particle swarm optimization (PSO) .Firstly, we define a scoring function which is evaluated on a set of candidate viewpoints of models.Apart from the amount of information, the features selected by the function can also embody the appetence of a viewpoint.Secondly, we use PSO to search the most appropriate weights of different classes of 3Dshapes so that the configuration of weights of features in the scoring function can be obtained.The results of comparative trial shows that viewpoints selected by our method share more common with the subjective tendency of most people and have no lack of geometrical information.