3D scanning of an entire structure is performed by integrating measurements obtained from multiple viewpoints. Some additional measurements are required if there are unobserved regions of the target structure. Currently, viewpoints are empirically selected and therefore inefficient selection could happen. In this research, we propose a viewpoint selection method by which measurements of unobserved regions are efficiently performed. The method uses the ray tracing to calculate the visibility of unobserved regions from each viewpoint candidate and then repeatedly selects the best viewpoints. We carried out an experiment to evaluate measurement positions selected by the proposed method.