摘要
为了能够及时检测到图像场景中的运动目标,提出一种基于混合高斯模型的运动目标检测方法。借助把图像的像素值看成是前景高斯分布和背景高斯分布的组合,进行了背景估计和自适应背景更新;通过对背景差分后的图像进行多目标分割,进行了多个运动目标的检测。实验发现:随着图像序列的背景不断变化,基于混合高斯模型算法能够准确估计出变化的背景,通过将场景图像和背景图像进行差分,检测到感兴趣的运动目标。
In order to detect moving object in image scene in time,this paper presents a moving object detection method based on mixture Gaussian model.We regard the image pixel value as the synthesized process of the foreground Gauss distribution and the background Gauss distribution,and realize the background estimation and adaptive background update.Through performing multiple object segmentation after background subtraction,we can successfully detect multiple moving objects.The experiments show that when the background of image sequence is changing,the mixture Gaussian model based algorithm can estimate changed background accurately,and through subtraction between scene image and background image,the interested moving object can be detected.The experimental results for real image sequences demonstrate this moving object detection algorithm based on mixture Gaussian model is efficient.
出处
《应用光学》
CAS
CSCD
北大核心
2010年第4期574-578,共5页
Journal of Applied Optics
关键词
运动目标
目标检测
混合高斯模型
背景差分
moving object
object detection
mixture Gaussian model
background subtraction