Edge-based object tracking for dynamic projection mapping

We propose an edge-based object tracking for realizing dynamic projection mapping. In order to achieve projection to moving targets, we have to obtain 3D posture of its targets in real time and with high accuracy. There are a lot of approaches which tackle this issues, and among these approaches a camera-based 3D pose estimation is suitable because this approach does not impair the appearance of targets. However, the approaches using depth information or feature points has trouble processing in real-time because of their large calculation costs. Therefore, we estimate 3D posture from 2D images by using the contour of the target object that we can obtain quickly by edge detection. And at this time, by using edge continuity without complex analysis of edges, we prevent the decline of the tracking accuracy caused by the extra edges detected from target's surface and surroundings. Consequently, we realize a real-time and high accuracy tracking for dynamic projection mapping applicable to various objects.