基于背景差法的运动目标检测

Video object detection based on background subtraction

视频序列图像中,视频分割的主要目的是要在视频序列中分割出具有意义运动对象实体.背景差法能够很好地从一段视频中提取出运动目标.可靠的背景图像的提取是该算法的关键.表述了一种新的背景提取算法,利用图像序列的灰度统计特性来提取背景图像,并利用Surendra背景更新算法根据每帧图像对背景进行更新已获得可靠的背景.然后,将当前帧与背景作差,并对差值图像进行适当处理,这样运动目标就能够被精确地提取出来.

For video serial images, the main objective of moving segmentation is to segment and extract representative body from the images. In order to extract the moving object better, the method of making difference between the current frame and the background is presented in this paper. Extraction of the stable background is the key problem of this algorithm. A new method to extract the background is presented in this paper. By using the gray characters of the video images, a good background can be extracted from the former N frames. Then use the Surendra algorithm to update the background according to the current frame. Finally a stable background can be extracted. When the background is extracted, the difference between current frame and background can be made. After some processes on the difference image, the moving object can be extracted aCCUr rately.