3D Eye Position Detection Using Supervised Descent Method
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional 3D display technologies require users to wear equipment like glasses, causing inconvenience, and there is a need for a method to determine 3D eye position information accurately and efficiently for enhanced user experience.
Innovation Solution
A method and apparatus that identify the eye area in a facial image, verify 2D features using a supervised descent method (SDM) model, and determine 3D position information by establishing a 3D target model based on these features, converting the information into a 3D coordinate system for adjusting or rendering 3D images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional 3D display technologies use glasses or helmets, then 3D image viewing is achieved, but user convenience deteriorates due to the need to wear equipment
Solution Approach 1:
The patent extracts the 3D viewing capability from the physical glasses/helmet equipment and transfers it to the display device itself through integrated eye position detection and image rendering systems, eliminating the need for external viewing aids
Solution Approach 2:
The display device performs self-service by using its own camera and processor to detect eye position and render 3D images tailored to each user's viewing angle, without requiring external 3D glasses or helmets
2Measurement precision
If 2D feature verification is performed without stationary state determination, then processing speed improves, but measurement precision deteriorates due to eye movement interference
Solution Approach 1:
The system performs preliminary determination of eye stationary state before executing the full 2D feature verification process, filtering out frames with eye movement to ensure measurement precision is maintained while avoiding unnecessary processing
Solution Approach 2:
The system dynamically adjusts processing by continuously monitoring eye position stability and selectively applying verification only when stationary conditions are met, optimizing the balance between precision and processing efficiency
3Measurement precision
If repetitive training with multiple features (HOG, LBP, SURF, ORB, Gabor, DCT) is performed, then measurement precision improves, but device complexity increases
Solution Approach 1:
The training process is segmented into multiple stages: initial training with approximate features (HOG, SURF, ORB) followed by subsequent training with precise features (LBP, Gabor, DCT), allowing the system to build up accuracy progressively without overwhelming complexity
Solution Approach 2:
The system changes training parameters by using different feature types at different training stages, starting with computationally simpler features and progressing to more precise features, thereby managing complexity while improving accuracy
Data Source
AI summary
A method of determining eye position information includes identifying an eye area in a facial image; verifying a two-dimensional (2D) feature in the eye area; and performing a determination operation including, determining a three-dimensional (3D) target model based on the 2D feature; and determining 3D position information based on the 3D target model.