›› 2012, Vol. 25 ›› Issue (6): 6-.

• 论文 • 上一篇    下一篇

一种基于全景视觉的运动目标检测方法

张忠民,李颖晶,贾艳丽   

  1. (哈尔滨工程大学 信息与通信工程学院,黑龙江 哈尔滨 150001)
  • 出版日期:2012-06-15 发布日期:2012-08-23
  • 作者简介:张忠民(1972—),男,副教授。研究方向:现代数字通信系统,数字信号处理。李颖晶(1985—),女,硕士研究生。研究方向:基于全景视觉的运动目标检测跟踪。贾艳丽(1986—),女,硕士研究生。研究方向:视频检测与跟踪。

A Method of Moving Target Detection Based on Omni-directional Vision

 ZHANG Zhong-Min, LI Ying-Jing, JIA Yan-Li   

  1. (School of Information and Communication Engineering,Harbin Engineering University,Harbin 150001,China)
  • Online:2012-06-15 Published:2012-08-23

摘要:

针对基于全景视觉系统采集的图像序列中运动目标检测的方法进行了研究,采用三帧差法和混合高斯模型相融合的算法对运动目标进行检测。不但解决了三帧差法无法检测出全部目标信息的缺点,而且还避免了背景减除法容易出现两个目标的假现象,在背景颜色和目标颜色相近时检测效果较好。实验表明,该算法能准确检测出运动目标,具有较好的实时性。

关键词: 目标检测, 全景视觉, 帧差法, 背景减除, 数学形态学

Abstract:

This essay proposes a method of moving target detection based on omni-directional vision,which detects a moving target by using the three frame difference and the mixture Gaussian model.Not only can it overcome the weakness of the three frame difference model that it fails detect all information of the target,but also it can avoid the false phenomenon of two objectives emerging by the background subtraction method.The effect is much better when the background color is close to the target color.Experimental results show that the proposed method can detect moving objects accurately in real time.

Key words: target detection;omni-directional vision;frame difference;background subtraction;mathematical morphology

中图分类号: 

  • TP391.41