Eye Glint Analysis for Facial Recognition Spoof Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional facial recognition systems are susceptible to spoofing attacks using high-quality printed photos or video displays, and 3D sensors for depth mapping are costly and not feasible for all devices.
Innovation Solution
A method involving controlled light pulses of varying intensities to capture glint reflections in the eye region, analyzing glint intensity curves, and comparing them with reference features to detect forgery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If 3D sensors are used to detect forgery by capturing depth maps, then the reliability of facial recognition is improved, but the device complexity and cost increase significantly
Solution Approach 1:
The patent uses a standard 2D camera to capture optical reflections (glints) from the eye region, creating a simplified copy of the depth information function without requiring actual 3D sensing hardware. The glint intensity patterns serve as a substitute for complex depth mapping, achieving forgery detection through optical property analysis rather than spatial reconstruction
Solution Approach 2:
The patent replaces the mechanical/optical 3D sensing system with a photometric analysis system using standard camera light pulses. Instead of measuring physical depth through structured light or time-of-flight, the system substitutes mechanical depth measurement with optical reflection intensity measurement, using glint patterns to infer surface properties and detect forgeries
2Ease of operation
If conventional 2D image capture is used for facial recognition, then the ease of operation is improved, but the reliability against spoofing attacks deteriorates
Solution Approach 1:
The patent employs periodic light pulses of varying intensities directed at the eye region, capturing a sequence of images at different illumination levels. This periodic illumination pattern creates distinct glint intensity variations that reveal optical properties of real eyes versus fake surfaces, enhancing anti-spoofing capability while maintaining simple 2D camera operation
Solution Approach 2:
The patent changes the illumination intensity parameter across multiple light pulses, capturing images at different brightness levels. By analyzing how glint intensity responds to these parameter changes, the system extracts optical property information that distinguishes real faces from forgeries, maintaining operational simplicity while improving reliability
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
Provides a cost-effective and efficient anti-spoofing mechanism by analyzing glint intensity patterns, enhancing security against printouts and displays, and integrating this method into vehicles and electronic devices.
Implementation Method 1
controlling a light-emitting device to emit light pulses according to the illumination pattern, and capturing with a camera a respective image of an eye region of a face during each emitted light pulse
Data Source
AI summary
A method for detecting forgery in facial recognition is disclosed that comprises selecting (110) an illumination pattern including at least three illumination intensities; controlling (120) a light-emitting device (220) to emit light pulses, according to the illumination pattern; capturing (130) with a camera (210) a respective image of an eye region (15) of a face (10) during each emitted light pulse; determining (140) whether a glint (20) is present in the eye region (15) of each captured image; measuring (150) a glint intensity of each determined glint (20); extracting (160) a respective numerical feature for each captured image, the numerical feature representing a curve of all measured glint intensities at the glint intensity of the respective image; and outputting (180) a signal that the face (10) is forged, if one or more of the extracted numerical features does not correspond to a reference numerical feature. Furthermore, a computer-readable medium storing instructions to perform the method, and a vehicle (1) comprising a processor (250) capable of performing such method.