NIR Material Spectroscopy for 3D Mask Face Spoof Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biometric facial recognition systems are vulnerable to 'spoofing' using 3D masks that mimic the general facial shape of authorized users, which existing technologies struggle to differentiate from real human faces.
Innovation Solution
The use of spectral characteristics of human facial features in near-infrared (NIR) images to authenticate liveness, involving NIR illumination and image sensors with RGB and NIR sensing elements, to distinguish between real human faces and fraudulent reproductions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional facial recognition systems are used, then ease of operation is improved, but reliability deteriorates due to vulnerability to spoofing attacks
Solution Approach 1:
The patent changes the spectral parameter domain by capturing facial images in the near-infrared spectrum rather than visible light. Human skin has distinct spectral reflectance characteristics in the NIR range that differ from mask materials, enabling the system to maintain ease of operation while improving reliability through spectral analysis of facial features
Solution Approach 2:
The patent introduces NIR illumination as an intermediary element between the light source and the facial recognition system. This intermediary enables the detection of spectral characteristics that are not visible to the human eye, allowing the system to distinguish between real faces and masks without complicating the user interface or operation
2Reliability
If spectral analysis in NIR images is implemented, then reliability is improved by distinguishing real faces from masks, but device complexity increases
Solution Approach 1:
The patent makes the imaging system multi-functional by using the same camera hardware for both visible light and near-infrared imaging. The sensor can operate in different spectral bands without requiring separate dedicated devices, thereby improving reliability through spectral analysis while minimizing the increase in device complexity
Solution Approach 2:
The patent combines visible light and near-infrared imaging capabilities into a single integrated system. By merging these functions and processing the NIR channel data alongside or instead of visible light data, the system achieves enhanced reliability for spoofing detection without proportionally increasing overall system complexity
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 effectively differentiates between live human faces and 3D masks by analyzing the spectral responses in NIR images, enhancing the security and reliability of biometric facial recognition systems.
Implementation Method 1
a first NIR image of a subject is acquired by illuminating the subject's face with a near-infrared (NIR) illuminator
Implementation Method 2
processing the first NIR image with a deep neural network to determine a first set of Gaussian distribution parameters corresponding to skin portions of the subject's face
Data Source
AI summary
A computer, including a processor and a memory, the memory including instructions to be executed by the processor to acquire a first image by illuminating a first object with a first light beam, segment the first image of the first object to determine regions that correspond to a first surface material and determine a first measure of pixel values in regions of the first image that correspond to the first surface material. The instructions include further instructions to perform a comparison of the first measure of pixel values to a second measure of pixel values determined from a second image of a second object, wherein the second image is previously acquired by illuminating the second object with a second light beam and when the comparison determines that the first measure is equal to the second measure of pixel values within a tolerance, determine that the first object and the second object are a same object.