电子学报 ›› 2015, Vol. 43 ›› Issue (7): 1444-1448.DOI: 10.3969/j.issn.0372-2112.2015.07.030

• 科研通信 • 上一篇    下一篇

基于局部特征的车载红外行人检测方法研究

王国华, 刘琼, 庄家俊   

  1. 华南理工大学软件学院, 广东广州 511400
  • 收稿日期:2013-10-28 修回日期:2014-07-07 出版日期:2015-07-25
    • 作者简介:
    • 王国华 男,1988年4月出生,广东阳春人.现为华南理工大学计算机学院硕博连读生,主要从事红外行人检测、模式识别等方面的研究. Email:w.guohuascut@gmail.com;刘琼 女,1959年3月出生,云南昆明人.现为华南理工大学软件学院教授、博士生导师,主要从事红外行人检测、模式识别等方面的研究. Email:liuqiong@scut.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61302121)

Method Research on Vehicular Infrared Pedestrian Detection Based on Local Features

WANG Guo-hua, LIU Qiong, ZHUANG Jia-jun   

  1. School of Software Engineering, South China University of Technology, Guangzhou, Guangdong 511400, China
  • Received:2013-10-28 Revised:2014-07-07 Online:2015-07-25 Published:2015-07-25

摘要:

车载红外行人检测在准确率和实时性方面存在多方挑战.文中基于行人头部、躯干成像与背景之间存在灰度分布差异,构建行人头部模型和躯干模型作为前端分类器,后端采用支持向量机(Support Vector Machine,SVM)进行分类;结合多帧校验和最近邻匹配跟踪行人.实验结果表明,检测时间基本持平,提高了检测准确率.

关键词: 红外视频, 行人检测, 头部模型, 躯干模型, 行人跟踪

Abstract:

There are lots of challenges in terms of precision and real-time performance in the detection of vehicular infrared pedestrian.This article established the pedestrians'head and torso models as the frond-end classifiers based on the brightness distribution difference between the pedestrians'head,torso and the background,and adopted the support vector machine (SVM) as the rear-end classifier;multi-frame check and nearest matching were combined to track the pedestrians.Experiment results show that the detection time is basically unchanged,and the detection accuracy have been improved.

Key words: infrared video, pedestrian detection, head model, torso model, pedestrian tracking

中图分类号: