Face Recognition Anti-Spoofing via PIV Displacement Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing biometric face recognition methods are not robust enough to reliably distinguish between real faces and two-dimensional images, especially in distributed IT environments where computational resources are limited and error-prone optical flow methods are inefficient.
Innovation Solution
A method using particle image velocimetry (PIV) principles to analyze displacement vector fields derived from two digital images of a face, without requiring pixel-level analysis or trained classifiers, by correlating image intensities and filtering out insignificant areas, allowing for efficient differentiation between real three-dimensional faces and two-dimensional photos.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical flow methods are used to distinguish real faces from photos, then measurement precision is improved, but device complexity and computational resources increase significantly
Solution Approach 1:
The patent divides the face image into multiple regions of interest (eyes, nose, mouth) and analyzes displacement vectors for each region separately. This segmentation allows the system to focus computational resources on key facial features rather than processing the entire image, thereby maintaining detection accuracy while reducing overall computational complexity.
Solution Approach 2:
The patent extracts only the essential displacement vector information from facial regions and compares these extracted features against reference patterns. By taking out only the critical motion characteristics rather than analyzing complete optical flow fields, the system achieves reliable detection with significantly reduced computational requirements.
2Measurement precision
If pixel-level analysis is performed to distinguish real faces from photos, then measurement precision is improved, but loss of time and computational resources increase
Solution Approach 1:
The patent applies partial action by analyzing only specific facial regions (eyes, nose, mouth) rather than performing pixel-level analysis across the entire face image. This selective approach processes a subset of pixels that contain the most discriminative information, achieving sufficient detection accuracy while dramatically reducing processing time and computational resource consumption.
3Measurement precision
If trained classifiers are used to distinguish real faces from photos, then measurement precision is improved, but device complexity and data transfer requirements increase
Solution Approach 1:
The patent employs rule-based comparison methods that do not require external trained classifiers. The system uses predefined geometric relationships and displacement patterns to make detection decisions, making the detection logic self-contained and eliminating the need for complex machine learning models. This approach maintains detection accuracy while significantly reducing system complexity and data transfer requirements.
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 provides a robust, computationally lightweight method to distinguish between real faces and photos using a small number of recordings, reducing errors and resource requirements, and enabling reliable identification in distributed IT systems.
Implementation Method 1
A method using particle image velocimetry (PIV) principles to analyze displacement vector fields derived from two digital images of a face
Data Source
Figure 1
Figure 2~3
Figure 4~6
AI summary
The method involves dividing two recordings of a face into a number of image components, where each image component comprises a number of pixels. The displacement of the individual image components from one of the recordings to the other recording is determined by a correlation process. A displacement vector field is generated from the displacement of the image components. The displacement vector field is analyzed to determine whether the recordings are made of a real face or from a two-dimensional (2D) diagram of the face. An independent claim is also included for a method for identifying a person in a distributed information technology (IT)-infrastructure such as cloud environment.