基于图像块的滞留物/移取物的检测方法  

An Abandoned / Removed Object Detection Method Based on Image Block

作　　者：李生金[1,2] 蒲宝明[2] 贺宝岳[1,2] 王维维[1,2] 

机构地区：[1]中国科学院研究生院,北京100039 [2]中国科学院沈阳计算技术研究所,沈阳110168

出　　处：《小型微型计算机系统》2014年第1期142-147,共6页Journal of Chinese Computer Systems

摘　　要：背景减法是智能视频监控系统中一种常用的运动目标检测方法.本文在均值化背景更新模型的基础上提出了一种滞留物或移取物检测方法.首先,将图像划分成大小相同的图像块,以图像块为研究对象建立均值背景更新模型;然后,采用背景减法,累加前景像素点对应记分板上的积分的方法,检测目标物的出现;最后,运用边缘匹配的方法识别出前景物体是滞留物还是移取物,并且根据识别结果采用不同策略更新目标物所在的背景图像块.实验结果表明,在复杂的室外多运动物的场景中,本文方法与传统基于运动目标跟踪的方法相比较,在正确率和CPU利用情况方面表现出较好的性能.Background subtraction is a widely used approach for detecting objects in video surveillance systems. The paper presents an abandoned or removed objects detection method based on the background subjection which uses the average value of frames as the background model. Firstly, the image is divided into the same size blocks. And the background model is provided according to the image blocks. Then once abandoned or removed objects appear in the scene, they will be detected via background subjection and accumulating the value in correspondence of foreground pixels on the scoreboard. Finally, we adopt an edge matching method to identify whether foreground objects are abandoned objects or removed objects and update background image blocks that contain foreground objects. Experiment results have indicated this method has more superior performance than the tradition methods based on moving object tracking in accuracy rate and CPU usage under the complex outdoor scenes with the moving multi-targets.

关 键 词：背景减法 图像块 滞留物 移取物 计分板 

分 类 号：TP312[自动化与计算机技术—计算机软件与理论]