Corneal Reflection Position Estimation for Eye Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pupil detection methods fail to accurately correct for corneal reflection positions when the head moves during image acquisition, leading to false images and inability to distinguish true from false corneal reflections.
Innovation Solution
A corneal reflection position estimation system that continuously acquires images of the eye using a camera with a light source, calculates movement vectors based on reference positions from multiple pupil images, and estimates the corneal reflection position in unilluminated images using these vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple image subtraction is used to detect corneal reflection, then the detection process is simple, but false corneal reflections cannot be distinguished from true corneal reflections when the head moves
Solution Approach 1:
The system performs position correction based on corneal reflection before subtracting the unilluminated image from the bright/dark pupil images. This preliminary position correction ensures that even when the head moves, the images are aligned properly, allowing true corneal reflections to be distinguished from false ones while maintaining a relatively simple detection process.
2Measurement precision
If position correction based on corneal reflection is performed, then measurement precision is improved, but the method cannot be applied when the head moves during image acquisition
Solution Approach 1:
The system performs position correction based on corneal reflection detected in the bright/dark pupil images before subtracting the unilluminated image. This preliminary correction is applied even when head movement occurs, enabling the method to adapt to dynamic conditions while maintaining measurement precision.
Solution Approach 2:
The system dynamically adjusts the position correction based on the detected corneal reflection positions in each image. When the head moves, the corneal reflection positions change, and the system adapts by using these changed positions for correction, making the method versatile under varying head movement conditions.
3Measurement precision
If multiple images are acquired to enable position correction during head movement, then measurement precision is improved, but the detection process becomes more complex
Solution Approach 1:
The system acquires the unilluminated image before acquiring the bright/dark pupil images. This preliminary acquisition of the unilluminated image allows it to be used for position correction in subsequent image processing, enabling accurate corneal reflection position estimation even when the head moves, while organizing the process in a manageable sequence.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate estimation and correction of corneal reflection positions in unilluminated images, improving the detection of true pupil positions by removing false reflections caused by head movement.
Implementation Method 1
a camera (10) equipped with a light source (13)
Implementation Method 2
a corneal reflection image is detected from a difference image acquired by subtracting the unilluminated image from the bright pupil image
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A corneal reflection position estimation system according to an embodiment includes: an image acquisition unit configured to continuously acquire an image of an eye of a subject by controlling a camera equipped with a light source, the image acquisition unit being configured to acquire a plurality of pupil images taken by using light from the light source and then acquire one unilluminated image taken without using light from the light source; and an estimation unit configured to calculate positions of corneal reflection or corneal sphere centers, which correspond to the plurality of pupil images, as reference positions, calculate a movement vector of the corneal reflection or the corneal sphere center based on a plurality of the reference positions, and estimate the corneal reflection position in the unilluminated image based on at least the movement vector.