面向远距离多用户虹膜识别的智能光场成像方法研究

Iris imaging is the first essential step of iris recognition. The limitations of current iris imaging method in terms of imaging distance, depth of field, posture and the number of users prevent wide applications of iris recognition technology. Therefore we propose a novel iris imaging method based on light-field imaging and intelligent vision algorithms to provide user-friendly iris image acquisition at a distance. Firstly light-field data is captured by the microlens array and multi-sensors optical system, and then computer vision methods for human-computer interaction such as head pose estimation, face detection and eye detection are integrated with image processing methods such as image stitching and digital refocusing to achieve large volume iris image acquisition of multiple subjects at a distance. The innovations of this project will estabilish an intelligent light-field imaging method for long-range iris recognition of multiple subjects and promote the development of iris recognition and opto-electonic engineering technology so that long-range iris recognition systems can be widely used in public security applications.

虹膜图像获取是虹膜识别的第一个关键步骤，现有虹膜获取方法在成像距离和景深、用户姿态和数量等方面存在诸多局限，制约了虹膜识别系统的大规模应用推广。本项目提出基于光场成像的远距离虹膜图像智能获取方法，通过构建微透镜阵列和多传感器光学系统获取高分辨率可供识别的四维光场数据，利用头部姿态估计、人脸和人眼检测定位技术实现智能人机交互，结合图像拼接、重对焦等图像处理方法共同实现远距离、大景深、高分辨、多用户的便捷式虹膜成像。本项目在虹膜图像信息获取源头创新，建立面向远距离多用户虹膜识别的智能计算型光场成像理论和方法，对于推动虹膜识别和光电工程学科发展具有重要意义，同时有效解决制约虹膜识别技术发展的距离、景深、用户数量等瓶颈问题，拓展虹膜识别系统的应用范围，满足我国关键领域对高精度身份识别技术的迫切需求。

本项目提出基于光场成像的远距离虹膜图像智能获取方法，研制成功国内首台面向虹膜成像的光场相机，提出了光场自动重对焦、光场超分辨算法，以及一系列基于光场成像技术的光场解码、深度估计、三维重建方法。本项目在虹膜图像信息获取源头创新，建立面向远距离多用户虹膜识别的智能计算型光场成像理论和方法，为全面了解和利用虹膜光场数据提供了一个基础实验平台，对提高虹膜生物特征识别的可靠性、鲁棒性和易用性具有重要的应用价值。本项目共在国际期刊发表论文2篇，国内期刊发表论文1篇，在国内外学术会议发表论文7篇，申请国内专利3项，其中授权1项。