Face Liveness Detection Using Multi-Angle Aspect Ratio Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing impersonation detection systems struggle to accurately differentiate between a real person and a photograph, leading to potential impersonation through image capture.
Innovation Solution
An impersonation detection system that captures multiple frames of a face at different angles, calculates the aspect ratio and three-dimensional similarity of facial features, and compares these ratios to detect impersonation using a plane surface object like a photograph with high accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If feature points such as eyes and mouth are used for authentication, then authentication can be performed with limited storage capacity, but impersonation using captured photographs becomes possible
Solution Approach 1:
The patent transitions from two-dimensional image comparison to three-dimensional spatial verification by capturing images at multiple angles and converting coordinates through 3D transformation. This dimensional enhancement allows the system to verify whether feature points lie on a plane surface, thereby detecting photograph-based impersonation while maintaining efficient storage of authentication data.
Solution Approach 2:
The system performs preliminary capture of images at multiple angles before authentication comparison. By pre-acquiring first and second images from different viewpoints and pre-converting coordinate systems, the authentication process can efficiently determine planarity without requiring excessive storage capacity for multiple high-resolution images.
2Reliability
If images are captured at different angles to detect impersonation, then photograph-based impersonation can be prevented, but data capacity and processing complexity increase
Solution Approach 1:
The patent extracts only the essential information needed for impersonation detection: feature point coordinates from multi-angle images. By converting these coordinates to a planar projection system and comparing whether they align on a plane, the system achieves high detection accuracy while avoiding the need to process and store entire multi-angle images, thus reducing computational complexity.
Solution Approach 2:
The system changes the parameter representation from raw pixel coordinates in multiple images to transformed 3D spatial coordinates. This parameter transformation enables the authentication process to work with compact numerical data rather than large image files, reducing both data capacity requirements and processing complexity while maintaining detection accuracy.
3Measurement precision
If feature point coordinates are converted to planar projection for comparison, then impersonation detection accuracy improves, but calculation complexity increases
Solution Approach 1:
The patent creates a virtual planar projection copy of the 3D feature point coordinates for comparison purposes. Instead of performing complex 3D geometry calculations with the original multi-angle images, the system generates simplified 2D projected coordinates that preserve the essential spatial relationships needed for impersonation detection, thereby reducing calculation complexity while maintaining precision.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Problem: Impersonation of a person using a plane surface object such as a photograph is detected with high accuracy. Solution: An impersonation detection system includes capturing means 11 that captures a plurality of frame images of an authentication target person at different angles, feature detection means 14 that detects an aspect ratio which is a ratio of a longitudinal length to a lateral length of a face of the authentication target person from each of the plurality of frame images captured by the capturing means 11, and impersonation detection means 14 that detects impersonation of a person by a plane surface object based on the aspect ratio of the face of the authentication target person captured at different angles.