改进视觉背景提取的运动目标检测算法  

Improved visual background extraction for moving target detection algorithm

作　　者：吴晗[1] 黄山[2] WU Han HAUNG Shan(College of Electrical Engineering and Information, Sichuan University, Chengdu 610065, China College of Computer Science, Sichuan University, Chengdu 610065, China)

机构地区：[1]四川大学电气信息学院,四川成都610065 [2]四川大学计算机学院,四川成都610065

出　　处：《计算机工程与设计》2017年第5期1282-1286,共5页Computer Engineering and Design

摘　　要：针对当前运动目标检测在实际应用中消除鬼影时间较长,对复杂多变的环境适应性不强等问题,提出一种改进视觉背景提取的算法。通过视频的前K帧图像中的奇数帧初始化背景模型和最大欧氏距离点的替换,加快鬼影的消除;将实时视频信息引入参考帧,借助参考帧的缓冲作用实现对背景的逐渐更新;参考帧和背景模型交替作用,充分利用像素点的空间信息和时间信息,获得真实可靠的背景图像。实验结果表明,该方法能够较快去除鬼影,大幅滤除缓慢运动的目标对背景的污染,消除由动态背景产生的误检,提高检测目标的准确率。For long time consumed when detecting moving target in practical application of eliminating ghost, and weak adapta-bility in the complicated and changeable environment, an improved visual background extraction algorithm was proposed. Through odd-numbered frames initialized background model in the pre-K-frame image and the replacement of the largest Eucli-dean distance point in the beginning of video, ghost elimination was speeded up. Real-time video information was taken into the reference frame, and reference frame buffer action was utilized to realize updating background gradually. The reference frame and the background model were functioned, and the pixels spatial domain information and the pixels time domain information were fully used, which obtained a reliable background image. Experimental results show that the proposed method can effectively get rid of the ghost, and it can filter influences of slow movement target on background, and eliminate the error detection caused by dynamic background, which improves the detection accuracy of target.

关 键 词：运动目标检测 背景模型 鬼影 参考帧 空间信息 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]