Eyeball Parameter Estimation via Geometric Axis Intersection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to stably and efficiently estimate the eyeball central position while accounting for varying eyeball radii, which are essential for accurate line of sight estimation in camera images.
Innovation Solution
An eyeball parameter estimating device and method that includes a head posture estimating unit, eyeball central position candidate setting unit, and eyeball parameter estimating unit, which calculate and refine the eyeball central position and radius by minimizing deviations of the point of gaze, pupil center, and eyeball center from a straight line, using feature points on the eyeball and accounting for individual differences in eyeball radius.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the eyeball is assumed as a single sphere and the center of the eyeball and the center of the pupil are assumed as two points for defining the line of sight, then the line of sight can be estimated without depending on positions of the camera and the illumination light source, but the eyeball central position cannot be directly detected from the camera image
Solution Approach 1:
The patent introduces an intermediary geometric relationship: the eyeball center is inferred as the intersection point of the visual axis (line from pupil center through fixation point) and the corneal surface normal at the corneal reflection point. This intermediary construction enables indirect detection of the eyeball center without direct observation, resolving the detection difficulty while maintaining independence from camera and illumination positions.
2Measurement precision
If various methods involving defining the line of sight are used, then line of sight estimation can be performed, but it is impossible to realize the estimation of the eyeball central position, in which calculation is stably converged at high speed while an eyeball radius varying between individuals is taken into account
Solution Approach 1:
The patent transforms the complex eyeball parameter estimation problem into a simpler geometric construction by changing the approach from iterative numerical optimization to direct geometric calculation. By defining the eyeball center as the intersection of the visual axis and corneal normal, and using the corneal reflection point to establish the visual axis direction, the method achieves both high precision and fast convergence without requiring iterative adjustment of individual eyeball radius parameters.
3Measurement precision
If the corneal reflection method is used, then a rotation angle of the eyeball can be calculated based on cornea reflection image, but the person is required to wear the camera or the like and thus has inconvenience
Solution Approach 1:
The patent uses a camera positioned at a distance to capture images of the person's face, creating a visual copy of the eye region. By analyzing the corneal reflection point and pupil center in this copied image, the system calculates the visual axis and eyeball center without requiring the person to wear any equipment. This copying approach maintains measurement precision while eliminating the wearability constraint.
Data Source
AI summary
A head posture estimating unit estimates, from a face image of a person photographed by a camera, a head posture of the person in a camera coordinate system. A head coordinate system eyeball central position candidate setting unit sets candidates of an eyeball central position in a head coordinate system based on two-point eyeball feature coordinates expressed in the head coordinate system. A camera coordinate system eyeball central position calculating unit calculates an eyeball central position ({right arrow over (a)}) in the camera coordinate system based on the head posture resulting from the estimating of the head posture estimating unit, the candidates of the eyeball central position set by the head coordinate system eyeball central position candidate setting unit, and a pupil central position detected from the face image. An eyeball parameter estimating unit estimates an eyeball central position (O) and an eyeball radius (r) in the head coordinate system based on the eyeball central position ({right arrow over (a)}) in the camera coordinate system.