Abstract: Existing viewpoint optimization algorithms that leverage geometric feature measurements are inefficient in generalization.A feature-adaptive viewpoint optimization algorithm for three-dimensional models is proposed in this paper.First, the mixed features of three-dimensional models are extracted.Then the matching rules between mixed features and viewpoint optimization are trained by means of AdaBoost classifier.In this way, the problem of viewpoint optimization is transformed into a classification problem.For a 3Dmodel, the constructed classifier can be employed to make an adaptive viewpoint selection.Experimental results show that the proposed algorithm can effectively characterize the structural features and details, and achieves better performance than previous approaches that employ a single model feature.