This research proposes a algorithm for the detection of 3D scanning direction in dental impression model. As 3D scanning industry grew in the latest, 3D scanner related medical technology is introduced to the field of Dentistry. To make a dental prosthesis, dental technicians have used dental impression models that are made of casting materials. But dental impression models are easily transformed and take up much space. 3D scanning operations are required in order to convert physical model to CAD model. Almost dental technicians have no CAD related expert knowledge. Thus when the physical model is scanned, intuitively they have to determine the scanning direction. If the dental impression model has simple shape, intuitional decision of user can be more efficient than the proposed algorithm. But dental impression model has complex shape and global interference, therefore some area that can't be scanned can exist. To distinguish between the area that can be scanned and the area that can't be scanned, users have to expend time and effort on finding the optimal scanning direction. The proposed algorithm detects virtual holes(the area that required scanning) using GPU parallel computing and contact detection. And we pick one direction among 250 directions considering the global interference. We are not consider marked directions, and all procedure work every scanning time.