融合背景减法和帧差法的运动目标检测  

Moving Object Detection Combining Background Subtraction and Frame Difference

作　　者：王奎奎[1] 玉振明[2] 

机构地区：[1]桂林电子科技大学信息与通信工程学院,广西桂林541000 [2]梧州学院,广西梧州543000

出　　处：《电视技术》2015年第24期94-99,共6页Video Engineering

基　　金：国家自然科学基金项目(61562074);广西自然科学基金项目(2013XNSFAA019323)

摘　　要：针对背景减法前景目标检测前期存在的"鬼影",通过统计图像序列中各像素点处像素值的频次,高频次的像素值为背景像素,获得了无运动目标的背景,消除了"鬼影";对于帧差法检测结果内部易产生"空洞",在前景判断时增加了新的判决条件,改善了"空洞"问题;对于传统背景减法检测结果存在"拖影"问题,通过背景减法和改进帧差法的检测结果进行逻辑与操作,并将"与"结果作为背景图像更新判决条件。实验结果表明,该算法可以获得精确的背景图像,并且解决了"拖影"、"鬼影"及"空洞"问题,满足实时检测的需求。To eliminate ＂ghost＂ before foreground target detection of background subtraction method, no-moving target background is obtained according to count the frequency of pixel values at each pixel point within image sequence. High-frequency pixel is considered as background pixel. Then the ＂ghost＂ is eliminated. To eliminate ＂holes＂ inside of detecting result got by frame differ- ence, a new decision criterion is added when judging foreground. The problem of ＂holes＂ is improved. For the problem of ＂image tail＂ of detected result got by traditional background subtraction, a logic ＂and＂ operation is clone according to the detection results of the background subtraction and improved frame difference, and the results is regarded as the decision criterion of background up- date. Experimental results show that the algorithm can obtain accurate background and solve ＂image tail＂, ＂ghost＂ and ＂holes＂, and it fits for the real-time detection.

关 键 词：背景初始化 背景减法 帧差法 背景更新 

分 类 号：TN911.73[电子电信—通信与信息系统]