一种改进的三帧差分运动目标实时检测算法  

Motion Target Real Time Detection Based on Improved Three Frame Difference

作　　者：张英[1] 车进[1] 周鹏[1] 

机构地区：[1]宁夏大学物理电气信息学院,银川750021

出　　处：《计算机测量与控制》2015年第6期2073-2075,2078,共4页Computer Measurement &Control

基　　金：国家自然科学基金项目(61162020)

摘　　要：针对传统三帧差分法提取的运动目标存在大量的噪声和空洞,提出一种改进的三帧差分运动目标实时检测算法;该算法采用Surendra背景提取算法提取有效背景,对视频流中连续三帧图像分别进行背景减除,得到的结果作为反馈对背景进行选择性更新,利用HSV颜色空间去除阴影后进行三帧差分,将差分结果进行"与"运行,通过将中间帧背景减除结果与"与"运算的结果进行"或"运算,这样可以得到运动目标的完整信息;实验结果表明,该算法能够快速、完整、准确地检测出运动目标,可有效应用于实时监控系统。According to the traditional three frame difference method to extract moving targets existing a lot of noise and empty, puts for-ward an improved three frame difference method for real--time detection of moving targets. The algorithm uses the Surendra algorithm to extract background. For three consecutive frames in video stream, we use background subtraction respectively. The obtained results as feed-back update the backgrounds. Using HSV color space removes the shadow and then we use three frames difference. The differential results are ＂and＂ operation. We use intermediate frame image with background subtraction result the ＂and＂ operation result on ＂or＂ operation. It can get complete information of the moving target. The experimental results show that the algorithm can be fast, complete and accurate to detect moving object, and it can be effective used for real--time monitoring system.

关 键 词：背景选择性更新 阴影去除 三帧差分 运动目标检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]