基于灰度投影运动估计的ViBe改进算法

Improvement of ViBe AlgorithmBased on Gray Projection Motion Estimation

传统的ViBe算法在目标检测中是比较常用的检测算法,该算法因为便于实现、运算效率高等优点在运动目标检测方面得到了广泛的应用。但是传统的ViBe算法在分离前背景时经常出现“鬼影现象”。针对以上问题提出了一种基于灰度投影运动估计的ViBe改进算法,首先用三帧差分法和ViBe算法对运动目标进行了实时检测,分离了前背景后,发现由于背景发生变化时对检测的前景目标图像有误检现象,为了去除运动背景的影响,采用了灰度投影运动估计的方法,用于估计两帧图像的平移参数和缩放参数,并选取一个参照帧的背景图像作为参考,最后根据将运动参数映射到当前输入的背景图像中,实现了背景的补偿,得到了真实的背景图像。实验结果表明,本文算法在对运动区域的目标进行检测时可避免鬼影、抗动态背景模型干扰等方面有很好的优势,并可以准确的在动态背景下检测出运动目标的位置信息。

The traditional ViBe algorithm is a commonly used detection algorithm in target detection.This algorithm has been widely used in moving target detection because of its advantages of easy implementation and high computational efficiency.However,the traditional ViBe algorithm often has a"ghost phenomenon"when separating the background.Aiming at the above problems,an improved ViBe algorithm based on gray projection motion estimation is proposed.Firstly,the moving target is detected in real time by three-frame difference method and ViBe algorithm.After the background is separated,it is found that the background is detected due to the change of background.The foreground image has a false detection phenomenon.In order to remove the influence of the motion background,a gray projection motion estimation method is used to estimate the translation parameters and scaling parameters of the two frames,and the background image of a reference frame is selected as a reference.Finally,according to the mapping of the motion parameters to the background image of the current input,the background compensation is realized,and the real background image is obtained.The experimental results show that the algorithm can avoid ghosting and anti-dynamic background when detecting the target of the motion region.Model interference has good advantages,and can accurately detect the position information of moving targets in dynamic background.