从图像视频重建人的3D形状与识别技术研究  

Research on Reconstruction and Recognition of Human 3D Shapes from Images and Videos

作　　者：张立武[1] 王蕊[2,5] 高红丽 田启川[4] 何荣祖 马凌慧 王蕊 Zhang Liwu;Wang Rui;Gao Hongli;Tian Qichuan;He Rongzu;Ma Linghui(Trusted Computing and Information Assurance Laboratory,Institute of Software,Chinese Academy of Sciences,Beijing 100190;State Key Laboratory of Information Security (Institute of Information Engineering,Chinese Academy of Sciences),Beijing 100093;College of Applied Sciences,Beijing University of Technology,Beijing 100124;School of Electrical and Information Engineering,Beijing University of Civil Engineering and Architecture,Beijing 100044;Shenzhen Institutes of Advanced Technology,Shenzhen,Guangdong 518055)

机构地区：[1]中国科学院软件研究所可信计算与信息保障实验室 [2]信息安全国家重点实验室(中国科学院信息工程研究所) [3]北京工业大学应用数理学院 [4]北京建筑大学电气与信息工程学院 [5]深圳先进技术研究院

出　　处：《信息安全研究》2019年第10期935-943,共9页Journal of Information Security Research

基　　金：北京高等学校高水平人才交叉培养“实培计划”项目;深圳市基础研究(自由探索)(JCYJ20180302145645821)

摘　　要：基于生物特征的身份识别前景光明,然而还存在诸多技术问题阻碍其应用:如比较流行的人脸识别技术还没有达到理想的精度,单一的生物特征也容易受遮挡、姿态等各种因素影响.提出了一种从2D图像/视频重建3D形状并加以识别的技术思路,利用大量植入的3D形状知识约束,结合分类决策算法分析,评估2D图像/视频,完成图像目标定位、3D形状重建与识别.此外,还对如何解决远程人脸识别中的“换脸攻击”进行了探讨.Biometric-based Authentication has a bright future. However, there are still many technical problems to be solved in its application: for example, face recognition technology has not achieved the desired accuracy, and recognition based on a single biometric is susceptible to various factors such as occlusion and pose. We proposed a technical idea of reconstructing 3D shapes from 2D images/videos and recognition. Using a large number of implanted 3D shape knowledge constraints, combined with classification decision algorithm to analyze and evaluate 2D images/videos to finish object detection and alignment, 3D shape reconstruction and recognition. In addition, we also discussed how to solve the “face-swap attack” in remote face recognition.

关 键 词：人脸识别 3D重建 2D图像 知识约束 面向对象 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]