【作者】 刘鹏；

【导师】 丁晓明；

【作者基本信息】 北京交通大学 ， 信号与信息处理， 2009， 硕士

【摘要】 随着当前社会的信息化、数字化、网络化的飞速发展,生物认证技术成为保障信息安全的一种重要手段。手指静脉网络与其它生物特征认证方式相比,具有不易受日常损伤、成像不易形变、受外界环境影响小和不易伪造等特点,日益成为一种备受关注的生物认证方式。本文针对手指静脉网络特征的采集与认证的关键技术进行了深入研究。主要的工作包括:1.设计并实现了一种基于红外光采集器的手指静脉图像采集装置的初级原型系统,为以后实现实用化的采集设备提供了初步的基础。2.提出了基于局部熵的反锐化掩膜红外手指静脉图像增强算法。有效改善了手指静脉图像中存在的血管清晰度不高,图像灰度分布不均匀的问题。3.首次提出将基于曲率的屋脊边缘提取算法用于手指静脉网络的构造,直接得到单像素的手指静脉网络骨架。与传统的先分离静脉图像再提取静脉网络骨架的算法相比,该算法简单,而且提高了静脉骨架提取的精度。在静脉骨架基础上采用最小二乘直线拟合法对手指静脉网络进行二次特征提取,得到静脉网络线特征。并通过构造稀疏矩阵得到了稳定的维数统一的特征向量。4.最后论文尝试了采用四元数模型进行多特征的融合,四元数模型由线段斜率,截距,中点横坐标,中点纵坐标四组特征向量构成。此特征融合算法简单,而目识别率高。更多还原

【Abstract】 With the rapid development of society informationize, digital and networking, biometric technology is considered as an important technology for information security. Compare with other biometric authentications, the finger vein recognition technology has many advantages. Firstly the finger vein is not easy to be daily damaged. Secondly it is hard to deform. And it is a living body recognition with interior characteristic. Therefore it is become a major concern.In this paper, the finger vein image collection and authentication are conduct for in-depth study. The main work includes:1. We have designed and implemented a primal finger vein image acquisition device which based on infrared light. It build a preliminary base for latter to make a ripe device.2. Proposes a promising finger vein image enhancement algorithm. Under the constraint that keeps original characteristic of finger vein, the method improves contrast of images. The uneven distribution of image grey of the image can be also resolved.3. Proposes the method based on surface curvature for roof edge detection to extract the vena framework. It can obtain one pixel framework from finger vein image directly. In traditional process the one pixel framework should be acquired after the vena segmented from original image. In contrast with the traditional process the proposed method is more simply and has high precision. After the vena framework has been obtained, the least-squares approximation beeline method is proposed to get the line feature. By constructing a sparse matrix we can get stable dimension of feature vector.4. Finally, this paper fuse four features by quaternion modulus, this algorithm is straightforward, and get high recognition rate.更多还原