摘要
人眼检测在人脸识别中起着非常重要的作用,其检测的准确性可以有效地提高人脸识别率。该文扩展了一种Haar特征,并基于该特征运用Adaboost算法对人眼进行准确检测。首先基于扩展的Haar特征运用Adaboost算法检测到眼睛和眉毛的粗轮廓,然后将粗轮廓的灰度图像制成模板,最后与眼睛和眉毛非常相似的图像样本进行匹配,从而准确检测到人眼。该方法有效地避免该由于眼睛和眉毛非常相似而引起的误判。
Eye detection plays a very important role in the face recognition. This paper expands a Haar features by using Adaboost algorithm to accurately detect the eye. By combining an extremely efficient classifier based on Adaboost algorithm, the wide outlines of eye and eyebrow are detected, and then the templates of these wide doutlines are formed. Finally, these templates are used to match the pictures with eyes and eyebrow. Result shows that the wrong judgement due to the similitude of the eyes and eyebrow can be avoided.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2010年第2期247-250,共4页
Journal of University of Electronic Science and Technology of China
基金
电子信息产业发展基金(信部运[2007]292)
