【作者】 李婷；

【导师】 种兰祥；

【作者基本信息】 西北大学 ， 电路与系统， 2017， 硕士

【摘要】 随着现代科技的飞速发展,人们对于身份安全认证的关注度逐步提高。手背静脉是人体的固有生物特征,由于手背静脉稳定的存在状态和非接触的采集方式使其在生物识别领域有着潜在的应用价值。手背静脉身份识别主要包括三个步骤:手背静脉图像采集与预处理、手背静脉特征提取和分类识别。针对上述步骤,本文做了以下工作:首先,在手背静脉图像采集方面,自行搭建了手背静脉采集装置,并通过该装置采集了实验室30人手背静脉图像并建立了手背静脉数据库用于算法验证。在预处理方面,改进了二值化算法和细化算法,最终得到了连贯的单像素宽的手背静脉骨架图像,为之后的特征提取奠定了基础。其次,在特征提取方面,本文分别提取了手背静脉图像的纹理特征和结构特征。在纹理特征方面,本文提出了基于小波分解的手背静脉图像HOG特征提取方法。对手背静脉灰度图像进行二级小波分解,在低频子带图的基础上提取HOG特征作为手背静脉图像的纹理特征,并将降维后的26维的纹理特征向量用于身份识别。在结构特征方面,本文提取并合并了手背静脉骨架图像的不变矩特征和拓扑结构特征组成了10维的结构特征向量用于身份识别。最后,本文通过信息融合技术将手背静脉图像的纹理特征和结构特征融合,组成36维的特征向量用于身份识别。较单一特征向量而言,融合后特征向量的识别率有了较大提高。更多还原

【Abstract】 Because of the rapid development of science and technology,people’s consciousness of identity security are gradually increased.Hand vein is one of biological characteristics of human body.The hand vein identification has a potential application value because of its stable state and collection of contactless way.Generally,hand vein recognition consists of three parts:hand vein image acquisition and image processing,feature extraction and classification recognition.As described above,the work done in this paper is as follows:Firstly,we built a hand vein image collection device and collected thirty kinds of hand vein images.We put these images into database for algorithm validation.In the aspect of image processing,we improves binarization algorithm and thinning algorithm.Image processing lays foundation for feature extraction.Secondly,in the aspect of feature extraction,the texture and structure feature of hand vein image are extracted respectively.A method based on wavelet decomposition is proposed as texture feature.A twenty-six-dimensional vector is composed by the HOG feature of the low frequency sub-band diagram.A ten-dimensional vector is used as structure feature including invariant moments and topological features.At last,the information fusion technology is used to combine the texture and structure features,and a thirty-six-dimensional vector is used for identification.Compared with single texture or structure features vector,the recognition rate of the feature vector after fusion is higher.更多还原