Abstract: An object segmentation algorithm based on background subtraction is presented according to the features of river surveillance video. First pixels are classified by the hue and intensity values in the HSI color space; then the background pixels are determined by a block-based algorithm; at last the moving objects are extracted with the background subtraction.

Key words: background subtraction, background reconstruction, motion detection, video object segmentation, visual surveillance

摘要： 针对内河航道监控视频的特点，提出一种基于背景差法的对象分割算法。首先在HSI颜色空间里利用像素的色调和亮度对其进行归类；然后利用基于块处理的方法确定背景像素，并在背景缓慢变化和急速变化时，采用定时和实时的背景重构方法进行背景更新；最后利用背景差提取运动对象。

关键词: 背景差, 背景重构, 运动检测, 视频对象分割, 视频监控