Eye Tracking Calibration via Polynomial Point Indexing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Eye tracking systems in wearable devices face calibration challenges due to lens pincushion distortion, which affects gaze tracking performance and requires accurate and reliable distortion modeling for proper function.
Innovation Solution
A hardware calibration method using a calibration rig that acquires optical target images, indexes image points, and applies a polynomial approximation function to predict image point locations, compensating for distortion, and writes calibration values to the eye tracking device.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a lens is used in the wearable device for eye tracking, then the eye tracking platform can capture eye images, but pincushion distortion occurs in the captured images especially when the eye moves far from the lens
Solution Approach 1:
The patent applies preliminary action by performing hardware calibration before actual eye tracking operations. The calibration process pre-determines the distortion characteristics of the lens and creates a mapping between distorted image coordinates and real-world coordinates. This preliminary calibration data is then used to correct distortion in real-time during eye tracking, ensuring accurate gaze measurement without requiring complex real-time distortion modeling.
2Measurement precision
If hardware calibration is performed to account for optical distortions, then gaze tracking accuracy is improved, but the calibration process requires a calibration rig and multiple image acquisitions
Solution Approach 1:
The patent applies universality by designing a calibration rig that can serve multiple purposes: it provides known geometric patterns for calibration, acts as a reference standard for verification, and can be integrated into the manufacturing process. The calibration pattern uses simple geometric shapes (circles, lines) that are easy to manufacture and provide sufficient information for determining lens distortion characteristics, making the calibration system universally applicable to different eye tracking implementations.
3Reliability
If distortion compensation is applied to improve image accuracy, then calibration reliability is enhanced, but the processing complexity increases due to optimization algorithms
Solution Approach 1:
The patent applies segmentation by dividing the calibration process into distinct stages: (1) capturing multiple images of the calibration pattern at different positions, (2) detecting features (circles, lines) in each image, (3) calculating distortion parameters from the detected features using optimization algorithms, and (4) applying the calculated distortion model to correct images. This segmentation allows the complex optimization to be performed offline during calibration, while real-time eye tracking uses the pre-computed distortion model, reducing real-time processing complexity.
Data Source
AI summary
The invention is related to a method for calibrating an eye tracking device within a head-mounted display (HMD) comprising the steps of acquiring with the HMD via an image sensor, at least one optical target image from an optical target, wherein the optical target contains image points in a pattern, indexing image points within the optical target image wherein the image points are indexed by, selecting a rigid region of the optical target image, assigning indices to image points within the rigid region, fitting a polynomial approximation function to at least one column and one row of the image points of the region, predicting the location of at least one image point using the fitted polynomial approximation function, assigning the predicted image point an index, inputting indexed image points into an optimization algorithm that calculates a hardware calibration of the HMD, and writing hardware calibration values calculated from the optimization algorithm to the HMD unit.