【Technical details】
We propose a projection mapping technology to overlay images on a dynamic surface (i.e., translating and/or rotating in a 3D space). To achieve accurate geometric registration of projection images to a dynamic surface, online pose and position measurement of the surface is required. We solve this issue by combining computer vision and full-color 3D printing technologies.
We print a projection surface out from a 3D printer with visual markers, which are also printed at the same time. We measure the markers using cameras to estimate the pose and position of the surface. The marker placement is optimized to realize robust estimation for arbitrary poses of the surface to the cameras. Furthermore, we diminish the markers by applying a radiometric compensation technique, which corrects each projected pixel color so that desired projected colors appear on the markers. Another advantage of our technique is that the markers can be printed on the surface with mechanical accuracy, while they were manually attached in conventional methods. Consequently, our technique has a less error in pose and position estimation than conventional methods.

Movies

【Exhibition content】
This demonstration shows a projection mapping for a dynamic surface. Projection images are accurately mapped by analyzing captured visual markers printed on the surface from a 3D printer. The projected images are attached onto the surface even when it is moved and/or rotated. Furthermore, the markers are diminished by projection, where projected colors are corrected pixel by pixel so that desired colors are appeared on the markers.