非受控环境中同质与异质三维人脸识别技术研究

At present, the mobility of large population of multiple races becomes more and more frequent, and all the countries in the world pay increasing attention to security, thus highlighting the importance of person identification. 3D data offer an alternative to deal with the unsolved challenges in 2D face recognition, e.g. variations in illumination and pose, and convey very high complementarity to 2D data. In particular, the advent of the devices capturing 4D (i.e. 3D video) data in real-time shows a good potential that 3D face recognition works in practical use. However, current 3D face recognition techniques subject to laboratory environment, and are not able to meet the requirements of complex scenarios. Therefore, it will provide great theoretical significance as well as large application value to thoroughly investigate the key issues in 3D face recognition under unconstrained conditions...This project will systematically study homogeneous and heterogeneous 3D face recognition in uncontrolled cases, e.g. freely moving faces when talking, walking, and continuously changing head pose, including 4D face data preprocessing (de-noising, face detection, landmarking, pose estimation, reconstruction using low quality models, etc.), only shape based 4D face recognition, textured 4D face recognition, 3D-2D heterogeneous face recognition, and 4D-3D heterogeneous face recognition. These contents cover all the critical problems of 3D face recognition in the real world, and extend the traditional static 3D face recognition to 4D homogeneous face recognition and 3D related heterogeneous face recognition. This research will comprehensively complete the theoretical framework of 3D face recognition which involves some issues in the 2D and 4D domain as well, and give basic but strong support to the application of 3D face recognition in the real condition.

当前多种族、大范围人员流动频繁，世界各国对安全的要求与日俱增，身份识别重要性凸显。三维数据为应对人脸识别中光照、姿态等挑战提供了新手段，且与二维数据完美互补，尤其实时四维(三维视频)采集设备的出现，为三维人脸识别走向实用展示了可能性。但三维人脸识别目前局限于实验室条件，尚不能满足复杂应用的需求。因此研究非受控环境中的三维人脸识别技术，具有重要的理论意义与应用价值。..本项目拟针对非限制条件下的同质与异质三维人脸识别展开研究，包括四维人脸数据预处理(涉及噪声去除、人脸检测、标志点定位、姿态估计、低质量重建等环节)、(全)四维人脸识别、四维三维异质人脸识别、三维二维异质人脸识别等关键问题，将传统静态三维人脸识别拓展到四维同质和与三维相关的异质人脸识别，涉及面向实际应用各个方面。本项目将全面完善以三维人脸识别为中心，并涵盖二维与四维人脸识别的理论框架，为三维人脸识别的实际应用提供有力的技术支撑。

当前多种族、大范围人员流动频繁，世界各国对安全关注与日俱增，身份识别作用凸显。三维数据为应对人脸识别中光照、姿态等挑战提供了新手段，且与二维数据完美互补，尤其实时三维采集设备的出现，为三维人脸识别走向实用提供了机遇。但三维人脸识别目前局限于实验室条件，尚不满足复杂应用的需求。因此研究非受控环境中的三维人脸识别技术，具有重要的理论意义与应用价值。.本项目围绕非受控环境中同质与异质三维人脸识别展开研究，应对其面临的姿态变化、外部遮挡、数据质量低等问题，涉及数据预处理、几何特征提取、形状序列表示、异质特征匹配、多模态特征融合等多个重要环节，提出了渐进式由粗到精的三维人脸标志点定位、基于快速轻量流形卷积的几何特征提取、基于频域信号增强的形状序列表示、面向便携式设备低质量深度图的三维人脸识别、基于深度典型相关分析与深度耦合谱回归的二维三维异质人脸识别等代表性方法。提出的方法在Lock3DFace、FRGC、BU-3DFE、BU-4DFE等国际权威数据库的测评上均达到了目前的先进水平，部分应用的结果处于领先地位，得到同行的积极评价。此外，项目建立了基于二维三维多模态数据的同质与异质人脸识别演示系统，实现了包含数据获取、形状分析到身份判别的完整处理框架。.在项目执行期间，发表学术论文28篇，其中IEEE TPAMI、IEEE TAC、ACM TOMM等国际重要期刊论文7篇（SCI检索/ SCI源刊），IEEE CVPR、IEEE FG、IEEE ICPR、IEEE ICIP、ACM MM、ACM ICMR、ACM ICMI、CCBR等主要国际和国内会议论文21篇（EI检索）。获2017 ICCV Workshop（AMFG）最佳论文奖、2017 CCBR 2017最佳学生论文奖等奖项；培养博士生3名，硕士生9名。