结合汗孔和细节点的高分辨率指纹匹配算法研究

With the development of automated fingerprint identification systems (AFIS) and people's growing demand for security, the existing minutiae-based AFIS are restricted at several crucial bottlenecks, such as the recognition accuracy, the anti-fake capability, and the sensitivity to frauds and attacks etc. To solve the above mentioned problems, this project proposes to develop high resolution fingerprint recognition techniques by considering both sweat pores and minutiaes. Sweat pores can only be extracted from high resolution fingerprint images captured by novel imaging sensor. They are found to be unique and stable enough for human identification. Our researches mainly focus on different kinds of algorithms with high accuracy and efficiency, for instance, sweat pore feature extraction, feature selection, feature alignment, feature matching, and feature fusion with minutiaes' algorithms and strategies. We plan to improve systems' recognition accuracy by combining more features. Meanwhile, such system is more robust to frauds due to the high complexity and large quantities of sweat pores. Sweat pores also can be used for liveness detection to strengthen system security. We finally aim to achieve a high resolution fingerprint identification system with high recognition accuracy and strong anti-fake capability. This work will promote the development of new generation fingerprint recognition techniques in our country.

随着自动指纹识别系统的不断推广，现有的单纯基于细节点的指纹识别技术已遇到瓶颈，如精度有待进一步提高、防伪能力差、容易受到欺骗或攻击等。为了解决上述问题，本项目提出结合汗孔和细节点的高分辨率指纹识别技术，借助新型图像采集传感器，获取高分辨率指纹图像，以捕获手指丰富的汗孔信息。手指皮肤表面汗孔分布具有个人特征，且相对稳定，可以用于身份识别。研究内容将围绕高精度且算法复杂度低的汗孔特征提取、特征选择、特征配准、和特征匹配算法，以及有效的汗孔特征和细节点特征融合匹配框架和匹配方法展开。通过扩大用于识别的特征集，提高识别系统的精度。同时，汗孔信息多且复杂，很难伪造，可增加系统的防伪能力。此外，汗孔特征还可以用来进行活体检测，以加强系统安全性。从而最终实现高精度、防伪能力强、及具有一定活体检测能力的高分辨率指纹识别系统，促进我国新一代指纹识别核心技术的发展。

本项目提出结合汗孔和细节点的高分辨率指纹识别技术，借助新型图像采集传感器，获取高分辨率指纹图像，以捕获手指丰富的汗孔信息。手指皮肤表面汗孔分布具有个人特征，且相对稳定，可以用于身份识别。研究内容主要包括（1）高分辨率指纹图像采集与质量评估：设计出了一款高分辨率指纹图像采集设备，提出了基于空间相关性度量的指纹图像质量评价方法；（2）高分辨率指纹图像中的汗孔提取：提出了基于参数分块自适应的汗孔提取方法、基于模型融合的汗孔提取算法，以及基于形态学滤波的汗孔提取算法，并通过实验证明了各种算法的有效性。（3）汗孔选择与汗孔匹配：提出了基于细节点三角剖分的汗孔匹配算法、基于脊线的汗孔匹配算法、基于确定性退火的汗孔匹配算法、以及基于图匹配的汗孔匹配算法，实验表明上述方法与已发表的论文结果相比，都具有一定的优势。同时，汗孔特征还可以用来进行活体检测，以加强系统安全性。在本项目执行期间出版英文专著1部，发表论文19篇，其中国际SCI检索国际期刊论文13篇，国际会议论文6篇。作为Program Chair组织召开生物特征识别领域的国际会议1次（2014, ICMB），参加国际/国内会议6人次，邀请境外专家来访并作学术报告6人次。