复杂环境下基于高斯混合模型的目标检测方法

Moving objects detection method based on Gaussian mixture model in complicated environment

在视频监控系统中,噪声和背景的扰动对于运动物体的检测都会产生显著影响。本文提出了一种改进的基于高斯混合模型的运动目标检测方法。该方法首先根据每个像素点在各高斯模型的时间尺度上的概率来更新权值得到背景图像;然后通过自适应阈值采用背景减除法检测出运动目标。实验表明该方法对于噪声和背景扰动都具有较强的鲁棒性。

In video surveillance systems, noise and shake of background can influence the detection of moving objects greatly. This paper proposed an improved moving objects detection method based on Gaussian mixture model. First, it based on the Probability of every pixel in the time-domain of each Gaussian model, which is used to update the Weight and the Background. And then in the method of background subtraction we can detect the moving object with the self -adaptive threshold. The experimental results the proposed method deals well with the influence of both noise and shake of background with strong robustness.