摘要
背景消除是智能视觉监控和自动目标识别与跟踪首先要解决的问题。采用混合高斯模型对背景图像进行建模,并应用图像的相关性与各点的置信度对背景模型的学习速率进行快速更新,完成背景模型的重建。与其他方法比较,该方法能够有效地对新背景进行快速的学习和适应,达到动态背景下运动目标实时跟踪的要求。
Background subtraction should be solved at first in intelligence surveillance and auto target recognition.A model of background image was constructed by adopting Gaussian mixture model,and it was reconstructed by updating the pixels learning rate rapidly by application of image correlation and pixel confidence.Compared to other methods,this method can make it efficient to learn and adapt new backgrounds rapidly,and achieve the requirment of real time moving target tracking under dynamic background.
出处
《鞍山科技大学学报》
2005年第3期239-242,共4页
Journal of Anshan University of Science and Technology
关键词
混合高斯模型
背景消除
自适应学习速率
背景重建
点置信度
gaussian mixture model
background subtraction
adaptive learning rate
background reconstruction
pixel confidence