Face Liveness Detection via Normal Map and Reflectance Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition systems are vulnerable to attacks using fake faces, such as 3D silicone or 3D print faces, and are affected by interference factors like posture changes, light reflection, and expression variations, necessitating improved face liveness detection methods.
Innovation Solution
A face liveness detection method that calculates a normal map and reflectance value map using incident light from different directions to capture 3D geometric and surface material information, determining the authenticity of a face image by considering these factors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional face recognition is used without liveness detection, then the system is simple and fast, but it is vulnerable to attacks using fake faces
Solution Approach 1:
The patent transitions from 2D face image analysis to 3D geometric information analysis by calculating normal maps that represent the three-dimensional surface orientation of the face. This dimensional transformation enables the system to detect fake faces by analyzing depth and surface geometry that cannot be captured in traditional 2D images.
Solution Approach 2:
The patent changes the detection parameters from simple motion features (blinking, mouth opening) to physical surface properties including normal map values, reflectance values, and curvature information. These parameter changes enable differentiation between real and fake faces based on material and geometric properties rather than just motion patterns.
2Reliability
If motion features like blinking or mouth opening are used for liveness detection, then the detection process is simple, but it cannot effectively resist fake face attacks
Solution Approach 1:
The patent moves beyond motion analysis in the temporal dimension to spatial dimension analysis by calculating normal maps that represent the 3D geometric structure of the face surface. This allows detection of structural differences between real and fake faces that are invisible in 2D motion-based approaches.
Solution Approach 2:
The patent replaces motion-based detection (mechanical system requiring user actions like blinking) with optical-geometric detection using normal map analysis. This substitution eliminates the need for user cooperation while maintaining high detection accuracy through physical surface property analysis.
3Measurement precision
If multiple light sources are used to capture 3D information, then the accuracy of geometric information improves, but the device complexity and cost increase
Solution Approach 1:
The patent makes the imaging device multifunctional by enabling it to capture both 2D face images and 3D geometric information (normal maps) using the same camera and light source setup. This universal approach eliminates the need for separate specialized hardware while achieving accurate geometric measurement.
Solution Approach 2:
The patent enables the imaging system to automatically calculate normal maps and extract geometric information from standard face images captured under controlled lighting conditions. The system performs self-calibration and self-processing, eliminating the need for external calibration equipment or complex additional hardware components.
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
Effectively resists various face liveness attacks by accurately distinguishing between live and fake faces, enhancing security in applications like finance, security, and attendance.
Implementation Method 1
calculates a normal map and reflectance value map using incident light in different directions to capture 3D geometric and surface material information
Implementation Method 2
calculates a normal map and reflectance value map using incident light in different directions
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A face liveness detection method, applied to the field of artificial intelligence. The method comprises: determining an initial face image and a set of face images corresponding to N illumination directions; determining N differential images according to the initial face image and the set of face images; generating a normal map and a reflectivity map according to the N differential images and the N illumination directions; and determining the face liveness detection result according to the N differential images, the normal map, and the reflectivity map. Also disclosed are a related apparatus, a device, and a storage medium. Three-dimensional geometric information and surface material information of the face image are taken into consideration simultaneously; therefore, the authenticity of the face image can be identified, and different face liveness attack approaches can be effectively resisted.