基于面部多模态生物特征信息融合的精准身份识别方法

本项目面向国家公共安全、“互联网+”行动、新一代人工智能战略重大需求和生物特征识别学科发展前沿，把握泛在信息网络（移动互联网、物联网等）为多模态生物特征图像获取与识别提供的发展机遇，系统研究多模态面部生物特征的精准成像、精准识别和精准鉴伪问题，提出基于多模态（人脸、虹膜、眼周、眼纹）、多层次（设备层、图像层、特征层、分数层）、多协同（数据和知识协同驱动机器学习模型、成像硬件和算法软件协同融合）信息融合策略的精准身份识别方法，解决弱光照、低质量、非配合、高动态等复杂条件下，通过计算成像和融合模型的协同创新突破现有生物特征识别的“感知盲区”和“性能瓶颈”，面向中国通用技术研究院的实战应用需求构建等错误率百万分之一的移动端和远距离场景精准身份识别验证系统，满足我国的国安、公安、金融、社保等关键领域对高精度、高可靠、高安全身份识别技术的迫切需求。

Accurate personal identification is a key problem for public security, Internet plus, artificial intelligence and biometrics research. And mobile Internet and Internet of Things provide great opportunities for acquisition and recognition of multi-modal biometrics. This project focuses on multi-biometrics research including accurate imaging, accurate recognition and accurate anti-spoofing. New ideas includes fusion of face, iris, periocular and eyeprint biometrics, multi-level information fusion strategies (device level, image level, feature level and score level), data and knowledge driven machine learning, fusion imaging device and algorithm for identity verification and liveness detection. The project aims to improve accuracy of personal identification under complex conditions such as weak illumination, low image quality, less-cooperative users and high dynamics using novel solutions of imaging and information fusion. We will develop mobile and long-range personal identification systems with extremely high accuracy (EER=0.0001%) for public security and finance applications.