基于OCT测量的皮下三维汗腺分割与形状表示

Biometrics authentication is widely used in the fields of national security and civil consumption. Optical coherence tomography (OCT) can measure subcutaneous 3D tissue, based on which biological features go to high anti-counterfeiting under-skin. Sweat gland is subcutaneous glandular of sweat pore, which can be considered as a potential biological feature with high recognition rate and anti-counterfeiting property, including the spatial distribution and individual shape. To investigate the spatial distribution, the sweat glands need to be separated. OCT measured sweat glands suffer from high speckle noise, intensity in-homogeneity and partial volume effect, which increase segmentation difficulty. This project proposes a nonlocal filter to remove strong speckle noise based on the similarity of tissues. Both spatial domain and frequency domain filtering for 3D similar groups are investigated. This project proposes a space-continuity and boundary smoothness enhanced fuzzy function, an adaptive cluster set up, to segment sweat glands from other biological tissues and get the spatial distribution of the sweat glands. To investigate the shape of the sweat gland, this project proposes to represent the shapes use spiral function based multiobjective polynomial curves, thereafter the spiral structure features are explored. So that this project provides theoretical basis and technical support for the research not only on the shape diversity of sweat glands but also on sweat gland comparison and analysis.

基于生物特征识别的身份认证在国家安全和消费领域广泛应用。光学相关层析成像(OCT)测量皮下三维组织结构，使生物特征得以向高防伪的皮下特征发展。汗腺是汗孔的皮下腺体，是潜在具有高识别率与高防伪性的生物特征，包括空间分布与个体形状两部分。为研究空间分布，需要对汗腺进行分割。OCT测量具有的高散斑噪声、强度不一致性和部分容积效应提高了分割的难度。本项目根据三维皮下组织重复性分布特点，提出基于非局部滤波算法，在时域与频域研究三维相似块组的滤波，消除强散斑噪声对汗腺分割的影响。本项目通过增强空间连续性、分割边界平滑性和模糊性的函数设计，基于分和处理的自适性类中心设置，实现汗腺的分割，得到汗腺的空间分布。为进一步研究汗腺的个体形状，本项目根据汗腺独特的单管螺旋结构，提出结合螺旋线方程的多目标多项式曲线拟合汗腺形状并研究形状特征，为汗腺个体形状多样性研究以及汗腺个体之间的比较分析提供理论基础与技术支持。

基于生物特征识别的身份认证在国家安全和消费领域广泛应用。光学相关层析成像(OCT)测量皮下三维组织结构，使生物特征得以向高防伪的皮下特征发展。汗腺是汗孔的皮下腺体，是潜在具有高识别率与高防伪性的生物特征。本项目自主搭建了一个1310nm的OCT测量系统，完成包含不同性别、年龄从20岁到65岁、不同工作类型的200余个手指，共1500余个OCT指尖体数据的三维采集。项目首先对这些体数据实现基于三维块匹配的降噪处理。其次从二维与三维两个角度对汗腺处理进行研究。通过自适应真皮深度，获取皮下汗孔图像，并提出了残差U-Net全卷积神经网络，实现对二维皮下汗孔的提取。提出一种融合双向卷积长短期记忆与全卷积神经网络的汗腺分割方法，充分利用相邻切片数据间的空间连续性，实现对三维汗腺的精确分割，最后对汗腺的管状结构进行建模。本项目不仅验证了汗腺作为指纹识别特征的有效性，同时通过对指纹膜的采集与对比，研究了汗腺的防伪性能，为汗腺作为皮下生物特征的研究提供理论基础与技术支持。