手指静脉血管网络重建与结构化识别理论问题研究

The finger-vein trait has been proven invaluable in personal identification since the finger-veins naturally hold the liveness and convenience as a biometric measure. The uniqueness of finger-vein comes from the complexity of vascular network in structure. However, the segmented vascular networks are always since the quality of the captured finger-vein images is very poor due to the heavy scattering of near-infrared light in tissue. This inevitably damages the uniqueness of the finger-vein pattern so that the reliability of finger-vein trait should be impaired greatly. Hence, in order to improve the performance of finger-vein recognition, a theoretical problem should be solved soundly about the real network reconstruction based on the extracted finger-vein fragments. In this project, we propose a novel finger-vein recognition theory based on vascular network reconstruction by making interdisciplinarity in human anatomy, biological fractal theory, granular computing and pattern recognition. The proposed strategies can effectively solve the problems which actually impair the reliability of finger-vein trait, such as vascular network reconstruction, hierarchical feature analysis and extraction, finger displacement. Hence, this project will change the environment of finger-vein recognition that relies heavily on finger-vein imaging quality, and promote the development of finger-vein recognition technology.

由于天然呈现活体性和便捷性，手指静脉在人的身份鉴别方面具有重要的科学价值。手指静脉的唯一性本质上是由血管网络的复杂特性所决定。但由于生物组织对近红外光具有高散射效应，手指静脉的成像质量往往较差，致使以图像为基础获得的静脉血管网络残缺破碎。这无疑会破坏手指静脉的唯一性，从而使其生物特征的可靠性大为降低。因此，能否以缺损血管为基础，获得完整的符合血管实际结构的血管网络就成了提高手指静脉识别性能的一个重要科学研究问题。本项目结合解剖学、生物分形学、粒计算和模式识别等相关理论提出了手指静脉血管网络重建下的新的手指静脉识别理论，可解决残缺血管网络的重建、层次化血管网络的特征分析与提取、手指的姿态变化等影响手指静脉可靠性的重要问题。本项目的开展将有助于手指静脉识别技术摆脱对静脉成像质量的严重依赖，必将进一步促进手指静脉识别技术深入发展。

由于生物组织对近红外光具有高散射效应，指静脉成像质量往往较差，血管网络易存在局部信息残缺的问题。本项目完成了项目计划书所设定的各项研究任务，从手指静脉血管图像的分形建模、网络修复与识别、鲁棒性特征表达、姿态失真校正等多个方面展开研究。结合生物分形学基本原理，实现了基于图像的指静脉血管网络分形特征提取，建立了指静脉血管网络分形修复识别模型。实现了基于轮廓线的指静脉图像姿态校正，并提出两种基于局部图结构的指静脉特征编码新方法，验证了对姿态变化的鲁棒性。依托本项目，还探索了基于深度模型的指静脉局部丢失信息恢复方法，构建了指静脉加权图模型并实现了基于图论的特征表达和识别，都取得了较好的结果。项目的研究为我们深入理解手指静脉血管网络生物学机理，修复成像中丢失的局部信息，提出更加鲁棒的指静脉特征表达和识别方法奠定了良好基础。