Abstract: An approach of background estimation is presented using characters of random variable moments and applied in moving object detection under static background.The program is optimized by graded and blocked method.The background is updated by zoning method using the result of detection.The computer simulation proves that background estimation based on high degree moment and background update in moving object detection is available.The same to improve the robust of background subtraction to the changing background in moving object detection.

Key words: high order statistic, background estimation, background update, moving object recognition

摘要： 利用随机变量的各阶矩的性质，构造了一种基于高阶统计量的背景估计方法，并将其应用于静态背景下的运动目标检测。文中采用多分辨率思想对程序进行优化。同时，在检测的过程中利用检测结果，采用分区域的方式对背景图像进行更新。计算机仿真实验表明，基于高阶统计量的背景估计及更新方法在应用于运动目标检测、提高背景差法在运动目标检测对背景环境变化的鲁棒性方面是有效的。

关键词: 高阶统计量（HOS）, 背景估计, 背景更新, 运动目标检测