上海交通大学学报(自然版) ›› 2012, Vol. 46 ›› Issue (12): 1920-1925.

• 自动化技术、计算机技术 • 上一篇    下一篇

基于改进混合高斯模型的交通流背景检测算法

吕复强1,王慧1,刘泓2   

  1. (1. 浙江大学 工业控制技术国家重点实验室, 杭州 310027;2. 浙江大学城市学院 信息与电气工程学院, 杭州 310015)
  • 收稿日期:2012-05-17 出版日期:2012-12-29 发布日期:2012-12-29
  • 基金资助:

    国家自然科学基金资助项目(50908204),浙江省教育厅重点科研项目(Z201018730)

     

Traffic Flow Background Detection Algorithm Based on Improved Gaussian Mixture Model

LV Fu-Qiang-1 , WANG  Hui-1, LIU  Hong-2   

  1. (1. State Key Laboratory of Industrial Control Technology, Zhejiang University, Hangzhou 310027; 2. Department of Information and Electronics, Zhejiang University City College, Hangzhou 310015, China)
  • Received:2012-05-17 Online:2012-12-29 Published:2012-12-29

摘要: 针对交通流视频中道路背景像素较为统一的特点,提出一种基于时空信息的双混合高斯模型背景检测算法.该算法先构造像素时间域混合高斯模型进行时间域的检测,并采用双重阈值分别判断前景与背景.当某像素无法准确判断时,针对该像素邻域构造空间域混合高斯模型,以空间域的检测结果代替时间域的检测.通过不同的交通流视频中的测试和比较,验证了所提出的算法能有效地融合像素自身的时间信息与像素间的空间信息,提高了检测初始阶段的鲁棒性,同时有效地解决了出现停车现象时的误检测问题.    

关键词: 交通流, 视频检测, 时空信息, 混合高斯模型, 决策融合

Abstract: Aiming at the uniform characteristics of background pixels in traffic flow video, a spatial-temporal dual Gaussian mixture model background detection algorithm was proposed. Firstly the temporal Gauss mixture model was built to detect the background in temporal domain, and a double-threshold method was introduced to detect the background and foreground. When the detection result was uncertain, the spatial Gauss mixture model was built, and the spatial detection result was used to replace the temporal detection result. The experimental comparisons with different traffic flow video demonstrate that the proposed algorithm in this paper improves robustness in the initial stage and decreases the detection fault also when parking phenomenon occurrs.

Key words: traffic flow, video detection, spatialtemporal information, Gaussian mixture model, decision fusion

中图分类号: