为了提高视频监控中运动目标分割的速度和准确度,研究并实现了一种基于高阶统计量HOS(HigherOrderStatistics)的分割算法。首先根据HOS假设检验处理帧差图,判定像素点是否属于运动区域,阈值通过灰度共生矩阵获得,考虑了背景纹理的慢变化。然后,用矩形框聚类法大致确定运动目标的范围,在该范围内使用形态运算法和首尾扫描法去除空洞。最后,使用模板相与法获得帧图像的运动目标模板,从原图像中分割运动区域。算法采用了由粗到精的分析策略,实验表明,是一种快速稳健的算法。 In order to improve the speed and accuracy of moving target segmentation in video surveillance, a segmentation algorithm based on Higher Order Statistics (HOS) is researched and implemented. First of all, the frame difference graph is processed according to the HOS hypothesis to determine whether the pixel belongs to the motion region. The threshold value is obtained through the gray level co-occurrence matrix, considering the slow change of the background texture. Then, a rectangular box clustering method is used to roughly determine the range of the moving target, and the morphological algorithm and the end-to-end scanning method are used to remove the void within this range. Finally, using the template phase and the method to get the moving target template of the frame image, the moving region is divided from the original image. The algorithm adopted from coarse to fine analysis strategy, experiments show that it is a fast and robust algorithm.