中文核心期刊
CSCD来源期刊
中国科技核心期刊
RCCSE中国核心学术期刊

重庆交通大学学报(自然科学版) ›› 2013, Vol. 32 ›› Issue (2): 365-0368.DOI: 10.3969 /j.issn.1674-0696.2013.02.41

• • 上一篇    

改进的混合高斯模型视频运动目标检测算法

魏建猛,陈松,庞首颜   

  1. 重庆交通大学信息科学与工程学院,重庆400074
  • 出版日期:2013-04-15 发布日期:2018-01-26
  • 作者简介:魏建猛( 1987—) ,男,山东枣庄人,硕士研究生,主要从事图形图像处理方面的研究。

Moving Objects Detection Algorithm in Video Sequence with Improved GMM

Wei Jianmeng,Chen Song,Pang Shouyan   

  1. School of Information Science & Engineering,Chongqing Jiaotong University,Chongqing 400074,China
  • Online:2013-04-15 Published:2018-01-26

摘要: 针对混合高斯模型背景建模在视频运动目标检测中的不足,提出了将混合高斯模型与三帧差分相结合来对视频中运动目标进行检测的算法。由混合高斯模型得到前景和背景,利用当前帧与混合高斯模型所得到的背景相减可以得到一个前景,使用三帧差分和边缘检测得到运动物体的精确轮廓,对此轮廓进行填充得到一个前景,将此三步前景进行运算得到最终的结果; 通过新的更新策略来快速地对背景进行建模,以像素点的稳定性来调整像素点的 更新速度,从而减少算法运算量,提高算法的运行速度。

关键词: 混合高斯模型, 三帧差分, 视频序列, 运动目标检测

Abstract: An algorithm for moving objects detection in video sequence based on Gaussian mixture model and three-frame infferencing is put forward to improve the deficiency of foreground detection based on Gaussian mixture model. Firstly,the oreground and background based on Gaussian mixture model is obtained. Then another foreground got by using the current frame minus the background is also obtained. Finally,the precious outline through three-frame differencing and edge detection can be obtained; meanwhile,a foreground is got by filling this outline. The final result is got by computing the foregrounds got from the above three steps. In addition,a new update strategy is used to model the background faster,the updating speed of model parameters is adjusted according to the stability of each pixel in frames to reduce the computational complexityand to improve the speed of the algorithm.

Key words: Gaussian mixture model, three-frame differencing, video sequence, moving object detection

中图分类号: