自适应背景更新及运动目标检测算法  

Adaptive Background Update and Moving Object Detection

作　　者：卢官明[1] 谢双[1] 

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003

出　　处：《南京邮电大学学报（自然科学版）》2013年第3期12-17,25,共7页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(61071167);江苏省科技支撑计划(工业)(BE2009157)资助项目

摘　　要：从视频中检测运动目标是智能视频监控应用中的一项关键技术。文中提出了一种基于区域的自适应背景更新及运动目标检测算法,首先使用高斯模型建立初始的静态背景图像,通过背景减法得到差值图像;然后使用自适应阈值对差值图像进行二值化,并利用形态学滤波和连通区域面积检测进行后处理,得到运动目标区域掩模;最后,根据运动目标区域掩模检测出运动目标,同时使用基于区域的自适应背景更新算法动态更新背景图像。实验结果表明,该算法能够自适应地对背景模型进行更新,对于背景的扰动、光线的渐变等带来的影响有很好的抑制作用,可以有效地检测出运动目标。Detecting moving objects from video is a key technique in intelligent video surveillance applications.This paper proposes an adaptive region-based background update and moving object detection algorithm.The initial stationary background is established by using Gaussian model and the difference image is obtained by background subtraction.Then,the binary processing of difference image is carried by using adaptive threshold,and the moving object region masks are obtained by the post processing including morphology filtering and area detection of connected regions.The moving objects are Finally detected and the background is updated.The experimental results show that the proposed algorithm can effectively deal with the problems of background disturbance,illumination change etc.and is more efficient than the algorithm based on pixel.

关 键 词：视频监控 运动目标检测 自适应背景更新 自适应阈值选取 

分 类 号：TN919.8[电子电信—通信与信息系统]