SBIR Phase I: A Face Identification System Using a Novel 3D Camera

Genex Technologies，Inc.的第一阶段小企业创新研究项目的主要目标是：是调查的可行性建设，第一次，一个自动的三维人脸识别（3D-FID）系统与复杂的三维人脸匹配能力，使用一种新的三维相机。 3D相机是一种电子3D成像设备，能够在一个快照中产生人脸的数字3D轮廓测量。 与仅使用2D灰度图像的传统Face-ID系统不同，Genex Technologies系统中的3D人脸图像提供了准确的几何信息，使Face-ID更加准确和高效。 人脸是一个三维（3D）物体，每个人脸都有其独特的3D表面轮廓。 人脸的3D特征和表面轮廓可以用于快速准确的人脸识别（Face-ID）。 然而，由于缺乏低成本的3D成像设备和有效的3D Face-ID算法，几乎所有现有的Face-ID系统仅使用2D灰度人脸图像作为其输入。 2D图像本质上容易受到光照条件和面部方向变化的影响，因为它们不能保留面部的3D几何结构。 这种基本限制极大地限制了当前Face-ID系统有效执行人脸识别功能的能力。 因此，现有Face-ID系统在现实应用中的匹配率非常低（低于90%）。 所提出的3D-FID系统的关键组件包括（1）能够实时获取2D和3D面部图像的新型3D相机;（2）基于高效3D面部识别算法的软件;（3）2D和3D面部图像数据库;以及（4）用户界面。 自动人脸识别系统的商业市场是相当大的，并且应用是众多的，包括在军事和民用领域。 安全和ATM机可以使用Face-ID来验证银行卡用户的身份，从而节省数百万美元的虚假持卡人身份。 计算机访问控制可以使用Face-ID系统而不是密码来授予计算机使用权限。 执法人员可以使用Face-ID系统根据逃犯的图像进行预选和锁定。 军方可以使用Face-ID系统来控制对禁区的访问。 贵重物品，如汽车或船只，可以使用3D-FID系统来识别其所有者。 低成本的3D相机也可用于许多其他应用，如制造零件检测、机器人、自动驾驶汽车和虚拟现实。 视频会议和交互式电视的快速发展进一步拓宽了3D成像技术的可能应用。

The primary objective of this Phase I Small Business Innovation Research project from Genex Technologies, Inc. is to investigate the feasibility of building, for the first time, an automatic 3D Face-ID (3D-FID) system with sophisticated 3D face matching capability using a novel 3D camera. What is meant by 3D camera is an electronic 3D imaging device that is able to produce digital 3D profile measurement of a human face in one snap shot. Unlike the conventional Face-ID system that use only 2D gray-scale images, the 3D face image available in Genex Technologies' system provides accurate geometric information that makes the Face-ID much more accurate and efficient. The human face is a three-dimensional (3D) object, and each face has its unique 3D surface profile. The 3D features and surface profile on a human face can be used for a rapid and accurate face identification (Face-ID). Due to lack of low-cost 3D imaging devices and effective 3D Face-ID algorithms, however, almost all existing Face-ID systems use only 2D gray-scale face images as their input. The 2D images are inherently vulnerable to changes in light condition and face orientation, because they cannot preserve the 3D geometric structure of a face. This fundamental restriction greatly limits the capability of current Face-ID systems in effectively performing face recognition functions. As a result, the match-rate for existing Face-ID systems in real-world applications is very low (below 90%). The key components of the proposed 3D-FID system include (1) a novel 3D camera that is able to acquire both 2D and 3D facial images in real-time; (2) software based on an efficient 3D face identification algorithm; (3) 2D and 3D facial image database; and (4) a user-interface. The commercial market for automatic face identification systems is sizable, and applications are numerous, including those in both military and civilian areas. Security and ATM machines can use Face-ID to verify the identity of bankcard users, saving millions of dollars in false claim of cardholdership. Computer access control can use a Face-ID system instead of password to grant authority for computer usage. Law enforcement agents can use a Face-ID system to preselect and nail down fugitives based on their images. Military can use the Face-ID system to control access to restricted areas. Valuable properties such as cars or boats could use the 3D-FID system to identify their owners. The low cost 3D camera can also be used for many other applications, such as manufactured parts inspection, robotics, autonomous vehicles, and virtual reality. The rapid advances in video-conferencing and interactive television further broaden possible applications of 3D imaging technologies.