基于LRP的自适应运动目标检测算法  

Adaptive Moving Object Detecting Based on LRP

作　　者：黄存东[1] 刘仁金[2] 杨思春[3] 

机构地区：[1]安徽国防科技职业学院,安徽六安237011 [2]皖西学院信息工程学院,安徽六安237012 [3]安徽工业大学计算机学院,安徽马鞍山243032

出　　处：《佳木斯大学学报（自然科学版）》2013年第2期253-255,259,共4页Journal of Jiamusi University：Natural Science Edition

基　　金：安徽省高校教学研究重点项目(20101689;20101686);安徽省自然科学基金项目(11040606M150);安徽省高校自然科学研究重点项目(KJ2011A048)

摘　　要：传统的运动目标检测算法主要基于像素值的统计模型,对于光照突变和噪声极为敏感.为此,提出了一种基于局部比率模式(LRP)的自适应运动目标检测算法.使用LRP描述视频图像序列中像素特征,通过自适应核密度估计对像素特征进行建模,提取出运动目标.实验结果表明,该算法适应光照变化,有良好的检测性能.Traditional moving object detecting algorithms are mainly based on statistical model by pixel value, which are extremely sensitive to illumination variance and noises. To resolve this problem, a novel adaptive moving object detecting method was proposed by model pixels with its local ratio pattern （LRP）. Pixel features in video image sequences were conducted by LRP, and the moving objects were extracted through kernel density estimation. The results of experiments on L2R databases demonstrate that the proposed has excellent detection accuracy on complex scenes. adaptive algorithm

关 键 词：目标检测 光照变化 局部比率模式 自适应核密度估计 带宽估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]