Customized 3D Eye Model for Six-Degree-of-Freedom Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current eye tracking methods are intrusive, uncomfortable, and unable to accurately recover all six degrees of freedom of eye movement, especially during corneal ablation surgery where the cornea's surface is altered, leading to systematic ablation errors due to misinterpretation of eye rotations as translations.
Innovation Solution
A method using a customized 3D eye model constructed during a model acquisition phase, aligned with current eye images to determine position and rotation coordinates, employing numerical fitting algorithms and texture information such as feature templates or gray value maps to track eye movements without relying on corneal reflections, allowing continuous tracking of all six degrees of freedom.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard non-intrusive techniques using corneal reflections are used to track eye movements, then it is possible to recover five degrees of freedom (three translations and two rotations), but it becomes impossible to accurately track eye movements after corneal ablation surgery when the cornea surface is no longer smooth and corneal reflections cannot be localized
Solution Approach 1:
The invention creates a 3D model (copy) of the eye's surface geometry and uses this model to track eye movements. Instead of relying on corneal reflections from the actual eye surface, the system projects the 3D eye model onto the camera image plane and compares it with the actual image. This copying approach allows accurate tracking even when the cornea surface is irregular, because the tracking is based on the pre-acquired 3D model rather than real-time corneal reflections.
2Device complexity
If only pupil detection in 2D camera image is used for eye tracking, then the system is simple and non-intrusive, but it is not possible to distinguish between eye rotation and eye translation
Solution Approach 1:
The invention transitions from 2D pupil detection to 3D eye model-based tracking. By acquiring the eye's surface geometry in three dimensions and projecting this 3D model onto the 2D camera image plane, the system can distinguish between rotations and translations. The 3D model contains depth information that allows the system to determine whether apparent pupil position changes are due to rotational movements or translational movements of the eye.
3Measurement precision
If head mounted cameras or cameras attached to head rest/chin rest are used for eye tracking, then all eye movements with respect to the camera can be interpreted as rotations and line of sight can be recovered by detecting the pupil, but the systems are intrusive and uncomfortable and require use of headrest, chin rest or helmet
Solution Approach 1:
The invention replaces the mechanical constraint system (headrest, chin rest, or helmet) with a computational approach. Instead of physically fixing the head to interpret all movements as rotations, the system uses a 3D eye model and image processing to mathematically determine eye pose. This substitution eliminates the need for intrusive mechanical devices while maintaining the ability to accurately recover line of sight and distinguish between rotational and translational movements.
Data Source
AI summary
A method of determining the spatial relationship of an eye (10) of a person with respect to a camera device (20) which provides images of the eye (10) comprises:a model acquisition phase in which a customized model of the eye (10) is constructed and a reference spatial relationship of the eye model with respect to the camera device (20) is determined using a reference image of the eye (10); anda tracking phase in which position and/or rotation coordinates of the eye (10) are determined by aligning the eye model to a current image of the eye (10).