基于分块差分与背景减除的运动检测

本文提出了一种基于分块的帧间差分与背景减除相结合的运动目标检测方法.该方法中,运动目标检测分两步进行,在粗分割阶段,首先把视频图像划分为多个子块,用帧间差分对运动图像各个子块进行检测,分离出背景区域和前景区域,背景区域用来构造和更新背景模型.一旦背景构造完成后,便可以对粗分割出的运动区域采用背景减除方法进行细分割,检测出精确的运动目标.使用帧间差分获取的背景区域构造和更新背景,能有效地把停止的运动目标融入背景,同时也能去除场景中突然运动的目标原来位置所留下"鬼影".实验结果表明,该方法克服帧间差分提取目标不精细缺点,解决了背景减除方法中背景构造更新的困难,能够对监控场景中运动目标进行精确检测,对光线变化、背景干扰不敏感,具有较好的鲁棒性和实时性.

Motion detection based on blocks frame difference and background subtraction

This paper presents an improved moving object detection method based on blocks frame difference and background subtraction. In coarse segment phase,the sequence image of video was divided into several blocks which was then been detected by interframe difference,background region,which is used to construct and upgrade the background model latterly,and foreground region was separated. Once the background was fulfilled,fine-grained segmentation can be carried on through background subtracting for the motion region been coarse segmented in above operation,so motion object can be accurately detected. The back ground region as well as the upgrading obtained by interframe difference can effectively take stopped motion object into background,and also can eliminate the "ghost" of the original position left by the sudden moved the object in the scene. Experiment results showed this method can solve the unsubtle problem in interframe difference object extraction,and also conquer the difficulty of background construction upgrading in background subtracting method. This solution can perform precise detection to motion object in monitoring scenes,and is immune to the light changing and background interference. It is of great robustness and real time operation.