基于数字全息显微（DHM）技术重建签名三维动态特征的个体识别新方法

This project creatively proposes the three-dimensional and dynamic attributions of handwriting in home and abroad, applies digital holography microscopy technology to realize the handwriting biometric recognition, and it’s expected to push handwriting biometric recognition forward into the development of four-dimensional direction. At present, the researches of handwriting biometric recognition are based on the measurements from static images which has big limitation. This project is based on the application requirement of biometric recognition, having the research strategy of digital holographic microscopy and Bayesian analysis, to explore the new method of individual identification of the three-dimensional and dynamic features of signature handwriting. This project is going to obtain the three-dimensional and dynamic features of signatures by digital holographic microscopy; to verify the specificity of three-dimensional and dynamic data with the application of statistical analysis methods; to realize the signature of individual identification with the employment of Bayesian analysis based on the construction of a database of 302000 signatures. This project focus to strengthen the theoretical basis of individual handwriting identification, to explore the combination of optical holographic technology, computer image processing techniques, statistical methods and forensic science, and intends to solve the key issues: extraction and the specificity validation of the three-dimensional and dynamic features of signatures, and stability and repeatability of individual identification system. The implementation of this project will contribute to promotion of interdisciplinary research and is helpful to deepen the understanding of multiple dimensions in handwriting features, moreover, it provides new theories, new technologies and new methodologies for biometric recognition.

本项目在国内外创新性的提出笔迹三维动态的属性，运用数字全息显微技术实现签名的个体识别，有望推动笔迹个体识别向4D方向发展。目前，笔迹个体识别是基于二维静态图像的测量进行的，存在较大的局限。本项目以个体识别的应用需求为导向，以数字全息显微技术和贝叶斯分析方法为研究策略，探索构建签名笔迹三维动态特征的个体识别新方法。项目运用数字全息显微技术重建签名的三维动态特征；运用统计学分析验证三维动态特征的特异性；在构建不少于302000个的签名数据库的基础上，运用贝叶斯分析方法实现签名的个体识别。项目着眼于增强笔迹个体识别理论基础，探索将光学数字全息技术、计算机图像处理技术、统计学方法和法庭科学技术相结合，拟解决签名三维动态特征的提取、特异性验证以及个体识别系统的稳定性和可重复性等关键问题。本项目的实施，有助于推动多学科的交叉研究，有助于深化笔迹特征的多维度认识，为个体识别提供新理论、新技术和新方法。

关于司法笔迹鉴定的争论已持续了几十年。目前笔迹识别研究总是聚焦于容易伪造的笔迹二维静态特征。本项目的研究解释离线笔迹的三维动态属性。运用三维和白光同步成像和图像处理，从304,200个签名数据库中获得三维和伪动态特征，数据库包括真实签名、代签签名、临摹签名和套摹签名，图像处理用来获得书写轨迹，从白光图像获得诸如笔画宽度、笔画弧度、骨架灰度和笔画灰度分布、骨架高度和笔画高度分布等伪动态特征；得益于三维图像和白光图像是同位同步成像，从三维图像上按照书写轨迹测量高度数据。三维伪动态特征的可视化和统计学分析探索笔迹形态学背后隐藏的与书写人无关的发现，帮助了解笔迹变化和个性的基本规律。相关系数用于表征笔迹之间的相似性，基于签名之间三维伪动态特征的比较测量带来了一些不同角度的观察。基于签名数据库中所有签名相对测量的ggpairs图显示稳定、显著的真实签名之间的相似性和真实签名与伪造签名之间的差异性。主成分分析的二维核密度图显示不同组签名（不同签名：不同的书写人，签名文本不同）的密度分布非常稳定。这不仅意味着本项目的签名数据库可以用作真实案例的背景数据，而且即便只有少量样本的真实案例，也可以使用这种方法。着眼于法庭科学应用，从签名数据库中随机选组一组签名，剩余的各组签名用来作为背景数据。评分似然比用来评估签名证据力度，性能评估显示等错误率0.000438。司法鉴定能力验证案例和真实案例的盲测正确率达100%，从而验证了本项目方法的稳定性和可重复性。