Optical Measurement Technology Precise Eye Tracking Through Deflectometry

Related Vendors

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

In a study, researchers developed an improved method to accurately track the gaze when wearing AR or VR glasses. This involves modern data processing. Over 40,000 surface points of the eye can be captured per camera image.

A research team at the Wyant College of Optical Sciences at the University of Arizona has developed an improved method for eye tracking that has the potential to significantly surpass existing eye-tracking technologies. The approach combines high-precision 3D imaging using deflectometry with computationally intensive algorithms from image processing. The goal is to capture significantly more surface data of the eye in real-time and thereby determine the gaze direction much more accurately than previously possible.

Advancements in Eye Tracking

Unlike common eye-tracking systems that work with a few infrared light sources, the new approach is based on a screen that displays structured light patterns. Each of the over one million pixels functions as a defined point light source. The reflected patterns on the eye's surface are analyzed to produce a dense 3D reconstruction of both the cornea and the sclera.

"By evaluating the deformation of the patterns, we obtain high-resolution topographic information of the eye," explains Dr. Jiazhang Wang, the study's lead author. In experiments with human subjects, an accuracy of 0.46 to 0.97 degrees was achieved—with artificial eye models, even up to 0.1 degrees.

The core of the development is the so-called stereo deflectometry. It uses two synchronized camera perspectives in combination with specially developed surface optimization algorithms. The advantage: The method does not rely on strong model assumptions about the eye shape—a clear plus in terms of variability between users. The generated data base also allows for a dense and precise surface reconstruction of the eye—a feature that is medically interesting beyond eye tracking. In the future, for example, eye diseases could be diagnosed or even compensated for in real-time.

Perspectives for Virtual Reality and Medical Technology

The integration of the technology into VR and AR systems is a declared goal of the team. Earlier work showed that structured light patterns could be directly integrated into the headset frame or embedded in digital content—also in the infrared range to avoid visually disturbing users. The technique creates very precise 3D images of the eye's surface, which could help diagnose and treat eye diseases in real-time.

There are already plans to commercialize the technology through Tech Launch Arizona, paving the way for new advancements in gaze tracking. Applications could include fields such as neuroscience and psychology, with artificial intelligence further enhancing the technology. The researchers hope to achieve the high precision they reached in model tests also in real-world applications, which would open up many new possibilities for next-generation eye tracking.

"Our goal is to achieve the accuracy of 0.1 degree that we reached in the model eye tests," says Florian Willomitzer, associate professor of optics and lead researcher of the study. "Our new method lays the foundation for next-generation eye-tracking technologies, including other applications such as neuroscience research and psychology," concludes Willomitzer. (heh)

Machine Authentication

Machine Operator Authentication - Why knowing "who" matters