ImageRefinement: Model-Free Marking Techniques
Geometric models of large areas are extremely useful in augmented reality, but are often either unavailable or incomplete. In this project, we proposed several different 3D user interface techniques in order to mark objects at long-ranges (10+ meters away). These techniques were all based upon a naive geometric technique, where users specified a ray pointing to the object from two close-by locations based upon the users’ head position and orientation. Through triangulation, the intersection of the two rays would define the position of the target object. However, this technique can have poor accuracy due to head tremors’ impacting the quality of the rays.
The first enhancement was to designate many more rays in a technique we call VectorCloud. This technique continually takes rays based upon the users’ head position and orientation. This minimizes the error from head tremors by averaging the intersections as more rays are taken.
The second enhancement was to improve the quality of the two collected rays from the naive geometric technique by having the user specify the object on an image taken by a camera attached the the user’s gaze position. An example of an image that is taken in an augmented reality implementation can be seen on the left. The user would move the target reticle to point at the target object. This ImageRefinement technique would then calculate the ray from the user’s head position to the position indicated on the image. This way head tremor is completely removed from the calculation of the intersection.
We did a preliminary study in virtual reality to evaluate the techniques with participants marking targets 15-75 meters away from the initial position. We then replicated the study in augmented reality in a real-world setting. This project was lead by my colleague Wallace Lages. I designed and implemented the ImageRefinement technique and implemented and performed the two studies. The results of the studies were published in ISMAR ’19 and can be seen here.
Lages, W., Li, Y., Lisle, L., Höllerer, T., & Bowman, D. (2019, October). Enhanced geometric techniques for point marking in model-free augmented reality. In 2019 IEEE International Symposium on Mixed and Augmented Reality (ISMAR) (pp. 301-309). IEEE. DOI: 10.1109/ISMAR.2019.00028