Enhanced Geometric Techniques for Point Marking in Model-Free Augmented Reality
PubDate: December 2019
Teams: Virginia Tech；University of California Santa
Writers: Wallace Lages; Yuan Li; Lee Lisle; Tobias Höllerer; Doug Bowman
Specifying points in three-dimensional (3D) space is an essential function in many augmented reality (AR) applications. When an environment model is not available, a straightforward solution is to perform geometric triangulation using two rays. However, naïve implementations suffer from low accuracy caused by technical limitations of AR devices and human motor constraints. To overcome these issues, we designed and evaluated two enhanced geometric techniques for 3D point marking. VectorCloud uses multiple rays to reduce the effects of pointing jitter, and ImageRefinement improves the accuracy by allowing users to refine the 3D direction of the two rays. We conducted studies to understand the characteristics of these techniques in both ecologically valid outdoor settings using a mobile AR display and in more controlled setting using virtual reality simulation. Our experiments demonstrate that both techniques improve the precision of 3D point marking, and that ImageRefinement is superior to VectorCloud overall. These results are particularly relevant in the design of mobile AR systems intended for use in large outdoor areas.