作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 图形图像处理 • 上一篇    下一篇

一种基于分裂高斯混合模型的背景建模算法

符 笛1,2,尹 东1,2,孙 涛2,3   

  1. (1. 中国科学技术大学电子工程与信息科学系,合肥230027; 2. 超精密控制与系统联合实验室,合肥230027;3. 中国科学院长春光学精密机械与物理研究所光电部,长春130033)
  • 收稿日期:2014-04-18 出版日期:2015-04-15 发布日期:2015-04-15
  • 作者简介:符 笛(1989 - ),男,硕士研究生,主研方向:智能信息处理;尹 东,副教授;孙 涛,副研究员。

A Background Modeling Algorithm Based on Splitting Gaussian Mixture Model

FU Di 1,2,YIN Dong 1,2,SUN Tao 2,3   

  1. (1. Department of Electronic Engineering and Information Science,University of Science and Technology of China,Hefei 230027, China;2. Joint Lab of Ultra Precision Control & System,Hefei 230027,China;3. Department of Photoelectronic,Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences,Changchun 130033,China)
  • Received:2014-04-18 Online:2015-04-15 Published:2015-04-15

摘要: 针对运动检测算法中目标空洞和环境噪声难以消除等问题,提出一种背景建模算法。将待处理图像通过三层高斯模糊滤镜,分层抽取图像信息,通过分层建立分裂高斯混合模型,进行建模并计算运动区域。将上层提取的低频信息以及建模后提取的粗块化运动目标,加入到下层的背景判定计算流程中,根据综合判定结果纠正下层错误的模型参数。在公共数据库中的实验结果表明,该算法在高效地去除了环境噪声的情况下,可保证所提取运动目标的准确性,并且对光照突变不敏感,有较好的检测率和较低的误检率。

关键词: 运动检测, 背景建模, 高斯混合模型, 背景差法, 分裂模型, 噪声消除

Abstract: Aiming at empty target and ambient noise,a background modeling algorithm is proposed. The image to be processed is filtered by three Gaussian blur filters in order to extract information stratified. Then,each layer builds splitting mixture Gaussian model group and computes change area. During processing, low-frequency information extracted from the upper layer is added to the modeling and gets contours of moving target which will join to the calculation of lower layer. The result tested on public video dataset indicates that the method ensures the accuracy of the extracted moving target when environmental noise is removed clearly. The method reduces the algorithm sensitivity for the scene illumination change and has a higher detection rate and lower false detection rate.

Key words: motion detection, background modeling, Gaussian Mixture Model ( GMM ), background subtraction method, splitting model, noise cancellation

中图分类号: