一种基于帧间差分与时空相关性分析的运动目标检测算法  

A Scheme for Moving Object Detection Based on Consecutive Frames Subtraction and Analysis of Space-time Relativity

作　　者：李惠松[1] 王小铭[1] 张玉霞[1] 

机构地区：[1]华南师范大学计算机学院,广州510631

出　　处：《计算机与数字工程》2007年第12期32-34,102,共4页Computer & Digital Engineering

摘　　要：分析现有复杂背景下运动目标检测与分割算法存在的不足,提出一种基于帧间差分与时空相关性分析的运动目标检测算法。结合帧图象中象素的空间相关性和帧间差分进行运动目标的分离,用序列图象中运动目标所具有的时间相关性来验证目标的真实性,剔除虚假的运动信息。实验结果表明,能较好地检测出目标,具有较好地抗干扰能力。On the basis of analyzing the deficiency of the existing schemes about moving object detection in the complex background, it gives a method of moving object detection based on the consecutive frames subtraction and analysis of space-time relativity. Firstly, it separates the moving object using the image’s space relativity and consecutive frames subtraction. Then time relativity is used to prove the actuality of the moving object. And the result shows that the method can detect the moving object well.

关 键 词：运动目标检测 帧间差分 时空相关性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]