基于三维镜框建模的眼镜虚拟试戴系统  

Glasses Virtual Try-on System Based on 3D Frame Modeling

作　　者：章锌栋 付东翔[1] ZHANG Xindong;FU Dongxiang(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《电子科技》2022年第9期52-57,共6页Electronic Science and Technology

基　　金：国家自然科学基金(61605114,61703277)。

摘　　要：在虚拟试戴系统中,虚拟眼镜需要精确地叠加到人脸图像上,这决定了虚拟试戴系统的体验感,其中的关键技术在于构建真实的眼镜模型以及快速准确地估计图像中人脸的三维空间姿态。针对该要求,文中提出了将三维镜框建模与基于人脸特征检测的头部空间姿态估计相结合的虚拟试戴技术,并实现了一个虚拟试戴系统。首先采用ERT级联回归定位人脸图像面部特征点;然后通过VTK可视化工具构建三维镜框,根据人脸特征点的信息和头部的旋转信息将处理后的镜框图像准确融合到人脸图像上;最后使用AFLW人脸数据库检验姿态估计的精确度。结果表明,该算法精度高、速度快,其在大角度、背景干扰物较多、光线条件差等复杂情况下仍能快速准确地实现三维多角度虚拟试戴,效果自然逼真,基本满足虚拟试戴技术的要求。In the virtual try-on system,virtual glasses need to be superimposed on the face image accurately,which determines the experience of the virtual try-on system.The key technology is to construct a real glasses model and quickly and accurately estimate the 3D face pose of the image.In response to this technical requirement,a virtual try-on technology combining 3D mirror frame modeling and head space pose estimation based on facial feature detection is proposed and a virtual try-on system is implemented.First,the method uses ERT cascade regression to locate the facial feature points of the face images.Then,the VTK visualization tool is adopted to construct a 3D mirror frame,and the processed mirror frame image is accurately fused to the face image according to the information of the face feature points and the rotation information of the head.Finally,the accuracy of the pose estimation method is tested using the AFLW data set.The results show that the algorithm has high precision and fast speed.It can still quickly and accurately realize three-dimensional and multi-angle virtual try-on under complicated conditions of large angles,many background interferences and poor light conditions,which basically meets the requirements of virtual try-on technology.

关 键 词：人脸特征检点 姿态估计 欧拉角 三维镜框建模 ERT级联回归 VTK可视化工具 消隐处理 图像合成 多角度虚拟试戴 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN99[自动化与计算机技术—计算机科学与技术]