Eyelid Detection Using 3D Face Pose Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing eyelid detection techniques fail to accurately detect eyelid positions due to red-eye phenomena and disturbances caused by glasses, leading to false detections.
Innovation Solution
An eyelid detection device that estimates the direction of a face by fitting feature points to a three-dimensional face model, limiting the detection range for upper and lower eyelids based on the face direction to exclude red-eye and glasses-related disturbances, and uses an upper and lower eyelid curve model fitted to an edge image for accurate detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge detection is performed on the entire face image to detect eyelid positions, then eyelid detection coverage is improved, but false detection increases due to red-eye phenomenon and glasses interference
Solution Approach 1:
The detection process is segmented into multiple stages: face region detection, red-eye region detection, glasses region detection, and eyelid detection. By dividing the image into distinct regions and processing each separately, the patent eliminates interference from red-eye and glasses while maintaining comprehensive eyelid detection coverage
Solution Approach 2:
The patent extracts and removes harmful elements (red-eye regions and glasses regions) from the detection process. By detecting these interfering elements separately and excluding them from the eyelid detection range, the system achieves accurate eyelid detection without false positives from red-eye phenomenon or glasses frames
2Area of stationary object
If the detection range is expanded to cover the entire face image, then the possibility of detecting eyelids is improved, but the influence of harmful factors such as red-eye and glasses increases
Solution Approach 1:
The patent segments the face image into multiple functional regions: face region, red-eye region, glasses region, and eyelid detection region. This segmentation allows the system to maintain a broad detection range while systematically excluding harmful factors from the final detection process
Solution Approach 2:
The patent converts the harmful red-eye phenomenon and glasses interference into useful detection targets. By detecting red-eye regions and glasses regions as separate entities, the system uses these previously harmful elements to define exclusion zones, thereby protecting the eyelid detection from false positives
Data Source
AI summary
An ECU 30 includes a face position and face feature point detection unit 32 that detects the feature points of the face of the driver, a face pose estimation unit 33 that fits the feature points of the face detected by the face position and face feature point detection unit 32 to a 3D face model to estimate the direction of the face of the driver, an eyelid range setting unit 34 that sets an upper eyelid presence range and a lower eyelid presence range including the positions of the upper and lower eyelids on the basis of the pose of the face estimated by the face pose estimation unit 33, and an eyelid detection unit 35 that detects the positions of the upper and lower eyelids in the upper eyelid presence range and the lower eyelid presence range set by the eyelid range setting unit 34.