计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (24): 188-190.DOI: 10.3778/j.issn.1002-8331.2008.24.057

• 图形、图像、模式识别 • 上一篇    下一篇

基于支持向量机的人眼检测

胡 涛,王家乐   

  1. 西安理工大学,西安 710048
  • 收稿日期:2007-10-26 修回日期:2008-01-24 出版日期:2008-08-21 发布日期:2008-08-21
  • 通讯作者: 胡 涛

Eye detection based on SVM

HU Tao,WANG Jia-le   

  1. Xi’an University of Technology,Xi’an 710048,China
  • Received:2007-10-26 Revised:2008-01-24 Online:2008-08-21 Published:2008-08-21
  • Contact: HU Tao

摘要: 人眼检测是计算机人脸识别的重要部分。复杂环境下人眼定位容易受到光照以及人不同姿态的影响。为了解决复杂环境下的人眼定位问题,采用基于支持向量机的人眼检测算法,首先对复杂环境下采集的不同人的人眼样本进行灰度化均衡以及小波变换,将变换结果表示成向量形式,运用序贯最小优化算法进行训练,得到一组支持向量,然后遍历待检测人脸图利用支持向量所构成的分类器进行人眼初检,最后根据先验知识完成信息融合,最终标定人眼。实验结果表明,该算法对各种复杂环境下的含人眼图像有普遍的适应性和有效性。

关键词: 人眼检测, 支持向量机, 序贯最小优化

Abstract: Eye detection is very important in face recognition .Eye detection on complicated environment is easily to affect by illumination and the posture of people.For solve the eye detection problem on complicated environment, in this paper a method of eye detection based on SVM is put forward.First,using gradation-rization balance and minor wave varies on the eye samples,and taking the result to the form of vector.This algorithm uses SMO method training the eye samples,which come from different people in complex background,and gets SV result.Then uses SVM for image detection,finally,it completes the windows merge.The experimental result indicates that,this algorithm has the common adaptability and validity to each kind of complex environment under including the human eye image.

Key words: eye-detection, Support Vector Machine(SVM), Sequential Minimal Optimization(SMO)