Abstract: This paper proposes a semi-automatic eyebrow recognition method based on Hash Graph based Semi-supervised Learning（HGSL） and Support Vector Machines（SVMs）.HGSL is presented to tackle the problem of time-consuming graph construction in Graph based Semi-supervised Learning method（GSL），for segmenting and extracting pure eyebrow images more efficiently in a semi-automatic way.The extracted pure eyebrow images are translated into feature vectors by Fourier transform and Gabor transform as well as principal component analysis，which are applied to training SVMs for recognition.On the BJUT eyebrow database，a series of experiments have been performed to analyze the effect of GSL and HGSL on eyebrow segmentation speed，and to summarize the influence of them together with feature and kernel selection on recognition rate.

Key words: eyebrow recognition, graph based semi-supervised learning, support vector machines, Fourier transform, principal component analysis

摘要： 提出了基于哈希图半监督学习和支持向量机的半自动眉毛识别方法。针对图半监督学习构图时间复杂度过高的缺点，提出了基于哈希图半监督学习的纯眉毛图像半自动提取方法。在纯眉毛图像的基础上通过傅里叶变换和Gabor变换及主成分分析提取纯眉毛图像的特征向量，用于支持向量机的训练和识别。在北工大眉毛数据库上，通过眉毛识别实验，分析了图半监督学习和哈希图半监督学习对提取纯眉毛图像速度的影响，并且总结了它们与特征向量和核函数的选择对识别率的影响。

关键词: 眉毛识别, 图半监督学习, 支持向量机, 傅里叶变换, 主成分分析