›› 2014, Vol. 27 ›› Issue (8): 22-.

• 论文 • 上一篇    下一篇

基于单目视觉的行人检测研究

韩永刚,赵巍凯   

  1. (西安电子科技大学 电子工程学院,陕西 西安 710071)
  • 出版日期:2014-08-15 发布日期:2014-07-21
  • 作者简介:韩永刚(1988—),男,硕士研究生。研究方向:图像处理与模式识别。E-mail:hanyonggangwudi@126.com。赵巍凯(1988—),男,硕士研究生。研究方向:软件开发和图像处理。

Algorithm for Pedestrian Detection Based on Monocular Vision

HAN Yonggang,ZHAO Weikai   

  1. (School of Electronic Engineering,Xidian University,Xi'an 710071,China)
  • Online:2014-08-15 Published:2014-07-21

摘要:

行人检测系统是目前先进驾驶辅助系统中直接面向行人的保护系统,可最大程度地减少行人所受到的伤害。纹理对称度特征是目前最直观且能够用于表征行人的特征。文中在采用基于纹理对称度特征方法提取感兴趣区域的基础上,提出了一种线检测的方法,可以有效地减少检测过程中阴影、树叶等小纹理对检测结果的影响。最后利用梯度方向直方图特征和支持向量机方法对感兴趣区域进行验证。试验结果表明,该方法在保证检测速度的前提下,可减少检测过程中的虚警和漏警情况。

关键词: 行人检测系统, 对称度特征, 线检测, 支持向量机

Abstract:

Pedestrian detection system is the pedestrian protection system in advanced driving-assistant system,which can effectively reduce pedestrian injury.Texture symmetry degree is currently the most proper feature that can describe pedestrians.A linear detection method is proposed by extracting the region of interest based on texture symmetry degree to effectively reduce the influence of small textures,such as shadow and leaves.The regions of interest are verified by histogram features of gradient direction and support vector machine method.Experiments show that the proposed method can significantly reduce false alarms and missing alarm without affecting the detecting speed.

Key words:  pedestrian detection system;symmetry feature;line detection;support vector machine

中图分类号: 

  • TP274+.5