Detecting rigid links between sensors for automatic sensor space alignment in virtual environments

Abstract

Simultaneous use of multiple sensor systems provides improved accuracy and tracking range compared to use of a single sensor system for virtual reality applications. However, calibration of multiple sensor technologies is non-trivial and at a minimum will require significant, and likely regular, user actioned calibration procedures. To enable ambient sensor calibration, we present techniques for automatically identifying relations between rigidly linked 6DoF and 3DoF sensors belonging to different sensor systems for body tracking. The techniques allow for subsequent automatic alignment of the sensor systems. Two techniques are presented, analysed in simulation for performance under varying noise and latency conditions, and are applied to two case studies. The first study identified sensors tracked by a gold standard rigid body tracker with one of six rigid bodies tracked by the first generation Kinect sensor with each sensor identified correctly in at least 76% of estimates. The second case study was an interactive version of the system that can detect a change in sensor configuration in 1–2 s and only requires movements of less than 15 cm or 90 ∘ . Our methods represent a key step in creating highly accessible multi-device 3D virtual environments.