人脸活体特征检测实验平台设计与实现

Design and realization of experimental platform for living feature detection of human face

在VS2013编程环境下,结合OpenCV设计了人脸活体特征检测实验平台。该平台由图像读取、人脸及五官检测、图像分块提取特征和机器学习部分组成。通过分析人脸真实图像与欺骗照片之间的纹理特征差异,实现对人脸真实性的判别。实验平台用于训练分类器的特征是Uniform Pattern LBP特征向量,将该特征应用于活体检测是本实验的创新之处。在该实验平台上对Replay-Attack数据库和实际创建的数据库进行实验,实验结果可以看出平台的准确率及运算速度等方面具有优良的性能。

In the environment of VS2013 programming,an experimental platform for living feature detection of human face is designed with OpenCV.This platform is composed of the image reading,face and five senses detection,image block extraction feature and machine learning.By analyzing the textural feature differences between real face images and deception photos,the discrimination of face authenticity is realized.The feature of the experiment platform used to train the classifier is the Uniform Pattern LBP eigenvector.The application of this feature to the living feature detection is an innovation in this experiment.On this platform,the experiments of the Replay-Attack database and the self-built database are carried out,and the experimental results show that the accuracy,operation speed,etc.,of the platform have excellent performance.