基于近红外图像特征的活体人脸检测

Face liveness detection based on near-infrared features

人脸识别技术目前虽然被广泛应用于生物特征识别等领域,但该技术仍存在被欺骗攻击的风险。为保障用户的合法权益,采用了一种基于近红外图像特征的活体人脸检测方法,以真假人脸在近红外光下成像存在纹理差异为依据,通过方向梯度直方图提取输入图像的局部特征,利用支持向量机进行二分类以判别真假人脸。最终在改进的CASIA NIR-VIS 2.0数据集和个人数据集上分别能得到97.2%和98.5%的分类准确率。

Although face recognition technology is widely used in biometrics and other fields,it still has the risk of being cheated.In order to protect the legitimate rights and interests of users,a live face detection method based on near-infrared image features is adopted.According to the texture difference of real and false faces in near-infrared imaging,the local features of input image are extracted by directional gradient histogram,and the support vector machine is used for binary classification to identify the real and false faces.Finally,97.2%and 98.5%classification accuracy can be obtained on the improved CASIA NIR-VIS 2.0 data set and personal data set respectively.