基于改进混合高斯模型的交通流背景检测算法  

Traffic Flow Background Detection Algorithm Based on Improved Gaussian Mixture Model

作　　者：吕复强[1] 王慧[1] 刘泓[2] 

机构地区：[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

出　　处：《上海交通大学学报》2012年第12期1920-1925,共6页Journal of Shanghai Jiaotong University

基　　金：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)

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.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.

关 键 词：交通流 视频检测 时空信息 混合高斯模型 决策融合 

分 类 号：U49[交通运输工程—交通运输规划与管理] TP391.4[交通运输工程—道路与铁道工程]