就如何从视频序列中分割出具有语义意义的运动对象 ,本文提出了一种自动的基于背景的运动对象分割算法 ,利用颜色、形状和灰度等特征对第一帧图像进行初步分割 ,然后根据帧间运动信息构造背景图像 ,最后以背景图像和帧差图像作为参考图像 ,对同一场景中的所有视频帧进行快速可靠的分割 。-On how to split out from the video sequences of moving objects with semantic meaning, this paper presents a context-based automatic segmentation of moving objects, using color, shape and gray-scale image features such as the initial fr a me of the first split, then inter-fr a me motion information structure background image, background image and finally the fr a me difference image as a reference image of the same scene in all the video fr a mes for fast and reliable segmentation.