基于特征点稀疏光流场的视频图像背景建模方法

An Background Modeling Method Based on Feature Point Sparse Optical Flow Field of the Video Image

针对利用高斯模型法进行背景建模所导致的背景图像出现目标阴影问题,本文提出了一种基于特征点稀疏光流场的视频图像背景建模方法。先利用角点检测找出目标特征点,然后利用光流场估计画出特征点光流并用矩形框框出,使大致的前景像素点包含在矩形框里,同时对处于矩形框外面各像素点的每帧图像进行背景建模,并采用Matlab编程软件对系统算法进行验证。验证结果表明,利用改进的混合高斯背景建模,可以去除背景灰度图像的阴影,使背景建模准确有效,并以此获得更精确的目标区域,而且该模型对基于背景减除法的目标检测算法更加有效,因此具有一定的实际应用价值。

For the use of gaussian model method for background modeling, as a result of the background image target shadow problems, this paper proposed a video images based on feature point sparse optical flow field in the background modeling method. Firstly, we use corner detection to find target feature points, and then draw the feature points which are estimated based on the optical flow field optical flow with rectangular box out. This makes the prospect of the general pixel be included in the rectangular box, at the same time, having image background modeling with each pixel of every frame outside the rectangular box, and we use Matlab software to verify this system algorithm. Verification results show that the improved gaussian mixture background model, can remove the background of the shadow of the gray image, make the background modeling more accurate and more effective, and obtain more accurate target area. The model of target detection algorithm based on background subtraction division is more effective, thus has certain actual application value.