EAPSI: Investigating methods for improving the alignment of augmented reality in see-through head-mounted displays

Moser Kenneth R, Meridian MS

Investigators

Abstract

Augmented Reality (AR) refers to the use of computer-generated content to enhance a person's view of the surrounding environment. Transparent head-mounted displays increase the effectiveness of these applications by providing the user a hands-free experience, while maintaining an unobstructed view of the world in front of them. An essential aspect of AR using a head-mounted display is proper alignment between the computer-generated elements and visible features in the world. This project will investigate the current methods used to correct misalignment in AR for see-through displays, and work toward reducing error and improving the accuracy of these techniques. This study will be conducted at the Nara Institute of Science and Technology (NAIST) in Japan in collaboration with Dr. Christian Sandor, a leading researcher in AR and Virtual Reality development. Results will aid AR developers seeking to use transparent display technologies for services ranging from the indication of bombs located around a soldier on the battlefield to overlaying MRI data onto a patient during an operation in order to help surgeons locate tumors. Augmented Reality systems typically employ one of two approaches to combine computer graphics with a real world view. Video See-Through AR, commonly employed on mobile devices, superimposes computer generated images over a camera feed. The second approach uses semi-transparent displays to overlay computer images directly onto a user's view of the world. Systems utilizing this see-through technology cannot directly access the view from the user's eye and therefore must use other methods to calibrate the placement of onscreen computed geometry. Common calibration procedures for these displays rely on user feedback to estimate the corrections needed. These methods, however, are prone to inaccuracies from system and environmental sources, such as tracking, human, and screen alignment error. This work will seek to measure the error contribution from individual sources and work toward the formulation of a general model for predicting and correcting misalignment errors automatically. This NSF EAPSI award is funded in collaboration with the Japan Society for the Promotion of Science.