基于特征学习的融合掌纹和手掌手背静脉信息的鲁棒身份认证方法研究

生物特征识别技术是一种有效的身份认证解决方案。特别地，在一些安全性要求高的重要应用场合，尤其需要高精度、高可靠性以及防伪性能好的鲁棒生物特征识别技术。近些年来，特征学习在很多研究领域取得了突破性的进展，获得了出色的识别效果。本项目提出基于特征学习的融合掌纹和手掌手背静脉信息的鲁棒身份认证方法研究，符合技术发展趋势。将首先开展掌纹、手掌静脉与手背静脉的新特征表示研究，重点构建判别性强的新局部特征描述子；然后开展面向局部特征描述子的特征学习方法研究，进一步提高识别精度，并在此基础上开展多模态识别算法研究；最后开展基于深度学习的掌纹、手掌静脉与手背静脉信息融合的识别算法研究。本项目将重点研究和构建一些特征提取、特征学习以及信息融合的新算法，并对它们进行理论探讨和分析，这些工作对于生物特征识别技术的基础理论研究将会起到一定的促进作用。本项目所研发的系统具有较好的应用前景，并能产生积极的社会影响。

Biometrics technology is one of the most effective solutions for personal authentication. Particularly, for some occasions with high security, robust biometrics solutions are urgently needed. In recent years, feature learning has made a breakthrough in several fields, achieving an excellent recognition performance. In this project, we propose a multimodal biometrics system simultaneously integrating palmprint recognition, palm vein recognition and dorsal hand vein recognition. At the same time, we will mainly explore feature learning methods for recognition. Firstly, research on new feature representations of palmprint, palm vein and dorsal hand vein will be carried out. We plan to design some new local descriptors with high discriminative power. Secondly, we will conduct research on descriptor learning to obtain a compact representation, using it to attain a better recognition performance. Lastly, deep learning methods will be investigated for multimodal recognition. Our project will further promote the research of palmprint recognition, palm vein recognition, dorsal hand vein recognition, and their multimodal recognition. At the same time, the multimodal system developed in this project has a promising market prospect.