复杂背景下的遗弃物检测  

Abandoned Objects Detection in Complex Background

作　　者：朱映映[1] 梁叶[1] 陈殷[1] 

机构地区：[1]深圳大学计算机与软件学院,广东深圳518060

出　　处：《小型微型计算机系统》2013年第5期1184-1188,共5页Journal of Chinese Computer Systems

基　　金：深港创新圈项目(ZYB200907060012A)资助;广东省自然科学基金项目(10351806001000000)资助;深圳市科技计划项目(JC200903120046A)资助

摘　　要：针对目前智能视频监控中复杂背景下的遗弃物检测存在较高的误检率,在结合混合高斯模型和三帧差分前景提取的基础上,研究了在较为复杂背景中对遗弃物事件进行自动检测和报警机制。首先,采用混合高斯模型和三帧差分结合的方法获得较为清晰的前景对象;其次,采用相交面积判定的追踪算法跟踪进入监控区域后静止不动的前景;再次,根据时间指标和距离指标判定暂时静止对象是否属于遗弃物;最后,对确认为遗弃物的对象进行报警。实验结果表明,该算法能有效的检测监控范围内的遗弃物,具有较高的检测精度和鲁棒性.In the field of intelligent video surveillance, abandoned objects detection in the complex background has faced high false rate. This paper designs an abandoned objects detection system based on the extraction of foreground with Mixture Gaussian Model and three -frame difference, which can obtain a mechanism for automatic abandoned objects detection and alarm in more complex background. At first, Mixture Gaussian Model and three - frame difference are used to obtain clearer foreground objects. Secondly, the intersection area of tracking algorithm is used to track the objects which enter the monitoring area and stop moving. Thirdly, According to time and distance to determine whether the temporarily stationary objects are abandoned objects or not. At last, if the objects are recognized as abandoned objects, the algorithm will give alarm. Experimental results show that the algorithm is robust enough to detect the abandoned objects effectively.

关 键 词：智能视频监控 遗弃物检测 混合高斯模型 三帧差分法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]