基于虚拟现实技术的光照变化人脸自动识别研究  

Research on automatic facial recognition with illumination changes based on virtual reality technology

作　　者：曾令菊 颜亮 ZENG Lingju;YAN Liang(Guilin University of Aerospace Technology,Guilin Guangxi 541004,China)

机构地区：[1]桂林航天工业学院,广西桂林541004

出　　处：《激光杂志》2023年第9期238-242,共5页Laser Journal

基　　金：广西科技厅科学技术委员会基金项目(No.2018KY0649)。

摘　　要：光照变化会模糊人脸特征,使得人脸识别准确性下降。针对这种问题,提出一种基于虚拟现实技术的光照变化人脸自动识别方法。采集光照变化下的二维人脸图像,利用虚拟现实技术重构人脸三维图像,提取人脸三维图像特征,计算试样本与人脸图像库中每张样本图像特征间的距离,并结合特征权重计算相似度实现人脸自动识别。实验结果表明:训练样本量为10时,所提方法的识别准确率达到98.1%。最低识别耗时为8.4 ms。不同光照等级下的F1值为8.8、9.2、9。同对照方法相比,所提方法的人脸识别准确率更高、识别耗时更短、综合评估能力更高。说明所提方法更适合处理光照变化的人脸识别问题。Illumination changes blur facial features and lead to reduce the accuracy of facial recognition.To solve this problem,an automatic facial recognition method based on virtual reality technology is proposed.The two-dimen-sional face image under the changing illumination is collected,the three-dimensional facial image is constructed by virtual reality technology,the three-dimensional image are extracted,,the distance between the test sample and the image features of each sample in the face image database is calculated,and the similarity is calculated by combining with the feature weights to realize automatic facial recognition.The experimental results show that the recognition accu-racy of the proposed method reaches 98.1%when the training sample size is 10.The minimum identification time is 8.4 ms.F1 values under different illumination levels are 8.8,9.2 and 9.Compared with the control method,the pro-posed method has higher accuracy,shorter recognition time and higher comprehensive evaluation ability.It shows that the proposed method is more suitable for facial recognition with changing illumination.

关 键 词：虚拟现实技术 光照变化 三维重构 特征提取 相似度 人脸自动识别 

分 类 号：TN209[电子电信—物理电子学]