摘要
针对运动目标检测问题中的背景建模问题,提出一种结合切比雪夫不等式和核密度估计的背景建模方法。首先利用样本均值与样本方差及切比雪夫不等式,快速计算各像素点属于前景和背景的概率,判别出前景点、背景点及可疑点。对可疑点再利用核密度估计方法,估计其属于前景与背景的概率密度来进行背景前景判别,最后通过设定阈值完成实时背景建模。实验结果证明,利用切比雪夫不等式能快速区分有明显特征的前景点与背景点,采用背景更新算法能得到理想的背景图像,降低了背景图像提取的误差,显著地提高了背景建模的速度。
To solve the background modeling problem in the motion target detection problem,a background modeling method that integrates Chebyshev inequality and kernel density estimation is proposed.First of all,depending on sample mean,sample variance and Chebyshev inequality,the method quickly calculates each pixel＇s probability whether it belongs to the foreground or the background,in order to classify it as a foreground point,a background point or a suspicious point.If it is a suspicious point,the kernel density estimation method is used to estimate its probability density whether it belongs to the foreground or the background in order to carry out background/foreground discrimination.Finally,by setting the threshold value the method completes the real-time background modeling.Experimental results prove that Chebyshev inequalities can quickly distinguish foreground or background points that bear significant characteristics,while by using the background updating algorithm it can obtain the ideal background image,reducing background image extraction mistakes,meanwhile remarkably speeding up background modeling.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第4期53-56,共4页
Computer Applications and Software
基金
国家自然科学基金项目(60273078)
关键词
切比雪夫不等式
可疑点
核密度估计
背景建模
Chebyshev inequality Suspicious points Kernel density estimation Backgroun modeling
