一种人脸防伪鉴别方法、装置、设备及存储介质
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JILIN UNIVERSITY
- Filing Date
- 2026-05-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing facial recognition technologies struggle to effectively distinguish between real and fake faces, especially when faced with attacks from high-precision simulation masks and 3D-printed masks, resulting in insufficient security. Furthermore, methods relying on single-modal feature acquisition and recognition or simple splicing of multi-modal data offer limited defensive capabilities.
A neural network model including a transformer encoder, CLIP model and classification head is used to process binocular images. By extracting semantic features, constructing disparity cost volume and fusing geometric embedding features, the authenticity of human faces can be identified.
It significantly improves the accuracy of face anti-counterfeiting identification, reduces false detection and false negative rates, and has high anti-counterfeiting capabilities against various forgery attacks. At the same time, it reduces identification costs, requires no special data collection equipment, and is suitable for various face recognition systems.
Smart Images
Figure CN122244929B_ABST