Anatomy-Constrained Gaze Estimation via 3D Coordinate Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gaze estimation technologies face inaccuracies due to image distortion caused by corneal surface refraction and are unable to account for torsional roll and individual non-symmetry of the eye, especially when head or device rotations occur, leading to computational errors and reduced accuracy in estimating point-of-gaze.
Innovation Solution
A system comprising an imaging module, a graphic processing module, and a central processing module that captures images, extracts features, determines head pose in a Cartesian coordinate system, and unprojects pupil and Iris contours to provide anatomically constrained gaze estimation, accounting for six degrees of freedom and individual eye symmetry.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simple perspective computations are used for pupil image processing, then the system is easy to operate, but measurement precision deteriorates due to image distortion from corneal refraction and sensor distortion
Solution Approach 1:
The patent transforms the 2D image coordinates of the pupil into 3D spatial coordinates by applying the inverse of the projection matrix, thereby changing the dimensional parameter to compensate for distortions. This allows accurate reconstruction of the pupil's true position and orientation in 3D space, resolving the measurement precision issue while maintaining computational feasibility
Solution Approach 2:
The patent replaces simple geometric perspective computations with a more sophisticated coordinate transformation system that uses projection matrices and their inverses. This substitution of the computational mechanism enables accurate correction of corneal refraction effects and sensor distortions, significantly improving measurement precision
2Device complexity
If existing gaze estimation methods are used, then the system has simple device structure, but reliability deteriorates when head or device rotations occur due to inability to account for torsional roll and individual non-symmetry
Solution Approach 1:
The patent transitions from 2D image plane analysis to 3D spatial coordinate analysis by applying the inverse projection matrix. This dimensional change enables the system to account for head rotations, torsional roll, and individual eye non-symmetry, significantly improving reliability while adding computational sophistication to the device
Solution Approach 2:
The patent explicitly accounts for individual non-symmetry of the eyes by processing each eye's pupil image independently through the coordinate transformation. This asymmetric treatment of potentially asymmetric biological structures improves reliability by adapting to individual anatomical variations rather than assuming symmetry
3Device complexity
If pupil centroid computation is used without distortion correction, then the computational process is simple, but measurement precision deteriorates due to non-linear distortion causing centroid displacement
Solution Approach 1:
The patent applies distortion correction through coordinate transformation before computing the pupil centroid. By preprocessing the image coordinates through the inverse projection matrix to obtain accurate 3D spatial coordinates, the subsequent centroid computation is performed on corrected data, eliminating displacement errors while adding a preliminary correction step to the process
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
This approach enhances the accuracy of gaze estimation by correcting for image distortions and accounting for head and device rotations, providing precise user-specific eye location and point-of-regard calculations in both 3D and 2D spaces.
Implementation Method 1
videooculography based upon the optical measurement of reflected light from the human eye
Implementation Method 2
reflected light is imaged onto a charge-injection (CID) or charge-coupled device (CCD) sensor array. The image of the eye is then electronically processed to determine the corneal reflection, the pupil centroid orientation
Data Source
AI summary
Anatomically-constrained gaze estimation providing a point of reference (PoR) in a 3D field and/or 2D plane, based on 6 DOF head pose constrained by the eyeball center being fixed in a common 3D coordinate system.