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

一种多模型协同的目标提取方法

冀续烨,陈 明,冯国富,赵海乐   

  1. (上海海洋大学信息学院,上海201306)
  • 出版日期:2015-05-15 发布日期:2015-05-15
  • 作者简介:冀续烨(1991 - ),男,硕士研究生,主研方向:图形图像处理;陈 明,教授;冯国富,副教授;赵海乐,硕士研究生。
  • 基金资助:
    国家“863”计划基金资助项目(2012AA101905)。

An Object Extraction Method of Multi-model Cooperation

JI Xuye,CHEN Ming,FENG Guofu,ZHAO Haile   

  1. (College of Information,Shanghai Ocean University,Shanghai 201306,China)
  • Online:2015-05-15 Published:2015-05-15

摘要: 固定摄像机目标提取多以高斯混合模型为背景模型,在检测运动缓慢、间歇停滞的目标时会出现前景目标 空洞的问题。为此,提出一种能够适应目标间歇停滞的多模型协同目标提取方法。采用高斯混合模型进行背景学 习,通过光线检测模型和场景状态检测模型协同控制背景适时更新,利用阴影检测模型剔除阴影。实验结果表明, 与KaewTraKulPong P 方法相比,该方法能较完整地提取到目标轮廓,且单帧处理时间较少。

关键词: 目标提取, 高斯混合模型, 光线检测模型, 场景状态检测模型, 阴影检测模型, 背景更新

Abstract: Gaussian Mixture Model(GMM) is adopted to solve foreground detection problems. However,GMM can not detect objects in which do not move in the scene. This paper proposes the multi-model cooperative method to detect foreground objects in complex scene. Under the assumption that the camera is fixed,it first uses the adaptive GMM to build a background which is updated by the light detection model and the scene detection model. A shadow detection model is also used in this paper at last. It mades a comparison with two algorithms. Experimental results show that this method can completely extract the object contour,and single frame processing time is less.

Key words: object extraction, Gaussian Mixture Model(GMM), light detection model, scene state detection model, shadow detection model, background update

中图分类号: