基于2D-3D人脸图像叠加比较的个体识别方法在汉族个体中可行性研究  

Feasibility study of individual identification based on superimposition of 2D-3D face images in Han individuals

作　　者：霍德民 张哲谌[2] 张晔 赵飞鸣 许文婷 杜猛[1] 张剑 王洪滔 周子皓 Huo Demin;Zhang Zhechen;Zhang Ye;Zhao Feiming;Xu Wenting;Du Meng;Zhang Jian;Wang Hongtao;Zhou Zihao(Institute of Criminal Science and Technology,Jiading Branch of Shanghai Public Security Bureau,Shanghai 201822,China;Department of Orthodontics,Shanghai Ninth People’s Hospital,Shanghai JiaoTong University School of Medicine,Shanghai 200011,China;School of Maritime Economics and Management of Dalian Maritime University,Liaoning,Dalian 116026,China;Beijing Entry-Exit Border Checkpoint,Beijing 100741,China)

机构地区：[1]上海市公安局嘉定分局刑科所,上海201822 [2]上海交通大学附属第九人民医院口腔正畸科,上海200011 [3]大连海事大学航运经济与管理学院,辽宁大连116026 [4]北京出入境边防检查总站,北京100741

出　　处：《中国法医学杂志》2023年第6期619-622,共4页Chinese Journal of Forensic Medicine

基　　金：上海市法医学重点实验室、司法部司法鉴定重点实验室开放基金资助项目(KF202210)。

摘　　要：目的 探讨基于2D-3D人脸图像叠加比较的个体识别方法在汉族个体中的可行性。方法 采集10位汉族个体的2D视频监控图像(包括正面、左侧面及右侧面图像)及高精度3D人脸模型,通过Autodesk 3ds Max 2018软件对3D人脸模型进行透视匹配处理以叠加至2D人脸图像上,并计算叠加后2D-3D人脸图像中11组对应特征点-点间距离的均值,将来自同一个体的2D-3D人脸图像叠加比较定义为匹配组,将来自不同个体的2D-3D人脸图像叠加比较定义为不匹配组。结果 整体而言,不论正面、左侧面或右侧面,匹配组与不匹配组间对应特征点-点间距离均值范围均不重叠(P <0.05)。结论 不重叠的均值范围初步表明本文所述的基于2D-3D人脸图像叠加比较的个体识别方法在汉族个体中具有一定可行性。Objective To explore the feasibility of individual identification based on the 2D-3D face image superimposition in Han individuals.Methods The 2D video surveillance images(including front,left and right side)and high-precision 3D face models of 10 Han individuals were collected,and Autodesk 3ds Max 2018 software was used to perform perspective matching on the 3D face models,and superimposed them on the 2D images,and the mean values of the distances between corresponding 11 feature points in the 2D-3D face images were calculated.The superimposition of 2D-3D face images from the same individual was defined as the matching group,and the superimposition of 2D-3D face images from different individuals was defined as the non-matching group.Results In general,the average distance ranges of the corresponding feature points between the matching group and the non-matching group did not overlap(P<0.05).Conclusion The non-overlapping mean range preliminarily indicates that the individual identification method based on the overlay comparison of 2D-3D face images described in this paper is feasible for Han individuals.

关 键 词：法医人类学 个体识别 3D人脸模型 2D-3D叠加 面部特征 

分 类 号：D919.6[医药卫生—法医学]