一种基于改进单高斯模型的运动目标提取方法

A Moving Object Extraction Method Based on Improved Single Gaussian Model

单高斯模型常用于室内场景或环境较为简单时的背景建模,适用于单模态背景情形,具有快速性和相对准确性。基于单高斯模型在运动目标检测中存在的缺点,提出了一种改进的单高斯背景建模方法。依据先验知识,为每一个像素点设定一个计数器来进行历史前背景点判别信息的统计,进而建立起一种"防欺骗"机制来防止背景点的误判,有效地抑制了传统单高斯模型中的拖影和前背景误判等现象。实验结果表明,改进方法在前景提取效果上有着显著的提高,背景形成更为准确.自适应和防干扰能力得到了增强,有效的防止了前背景点的误判。

In consideration of the shortcomings a single Gaussian model has in detecting moving objects. A modified single Gaussian modeling method is proposed. According to the prior knowledge, it establishes a counter for every pixel in the background-image to avoid misjudgments. Experiments have proved it effective and practica- ble in background-image misjudgments occurring in the traditional single Gaussian model.