研究目的
To propose a PAD method based on the medium wave infrared (MWIR) polarization characteristics of the surface material for countering a flexible 3D silicone mask presentation attack.
研究成果
Polarization-based MWIR imaging is more suitable for the study of the 3D silicone face mask PAD method than conventional MWIR imaging. The PAD method displays a certain robustness in the detection of facial temperature changes.
研究不足
The amount of data collected is not large enough to use deep learning methods due to the over-fitting of the network. Future works will expand the data amount to develop more advanced deep learning methods.
1：Experimental Design and Method Selection
A polarization MWIR imaging system is designed and built for face spoofing detection, utilizing the polarization characteristics of the surface material. The system captures facial images at four polarization angles to calculate the degree of linear polarization (DoLP).
2：Sample Selection and Data Sources
A sample database of real face images and 3D face mask images is constructed. Subjects are asked to sit approximately 220 cm away from the imaging system, facing the camera without wearing eyeglasses.
3：List of Experimental Equipment and Materials
MWIR camera with a resolution of 320 * 256, metal wire grid polarizer, optical experiment platform, and several other polarization device accessories.
4：Experimental Procedures and Operational Workflow
By rotating the polarizer, MWIR intensity images of four polarization angles are captured. Stokes parameters S0, S1, and S2 are calculated to obtain MWIR polarization images. A gradient amplitude feature extraction method is designed based on the DoLP of infrared radiation in different regions of facial images.
5：Data Analysis Methods
The gradient amplitude feature vectors are inputted into an SVM classifier for training and classification. A seven-fold cross-validation method is used to evaluate the PAD performance.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
