スダンシュ クマール セムワル, 大谷淳
Scan & Track: An Active Space Indexing System for
Unencumbered 3D Tracking in Virtual Environments
Abstract:We present a new method for unencumbered tracking of participants in a virtual environment using multiple cameras. During preprocessing, the system uses three cameras to record a planar-slice containing a simple pattern arranged on a regular grid. The planar slice is then physically moved at regular intervals and the corresponding camera-images are stored. In this manner, active-space or the 3D-space between these parallel slices is scanned. In this manner, we scan the active-space. The stored camera images are processed to create an interactive, active-space indexing mechanism which maps 3D points in active-space to the corresponding projections on the camera-images. The active-space indexing mechanism allows an operator to specify poses by specifying significant points on the camera-images of a participant. These significant points could also be automatically generated by analyzing the camera-images.
In this paper, we first discuss the Scan & Track system. Later, details of the active-space indexing method and results are presented. Our method avoids the complicated camera calibration operations, and is robust as the distortions due to camera projection are automatically avoided. In addition, the system is scalable, as active-indexing for the same 3D-space could be developed for both the low and the high-end systems. In addition, the active-space is also scalable. These qualities make the Scan & Track system ideal for future VR applications.
Key Words: 3D Motion Tracking, Multiple Cameras, Unencumbered VEs.