基于Davinci-DM6467的高斯混合模型算法的实现  

Realization of gaussian mixture model algorithm based on Davinci-DM6467

作　　者：刘德方[1] 王戴木[1] 邓明[1] 陈静[1] 赵正平[1] 

机构地区：[1]阜阳师范学院计算机与信息学院,安徽阜阳236037

出　　处：《阜阳师范学院学报（自然科学版）》2012年第2期69-72,76,共5页Journal of Fuyang Normal University(Natural Science)

基　　金：阜阳师范学院产学研项目(2010CXY04);阜阳师范学院产学研重点项目(KJ2012A218);安徽省教育厅自然科学研究项目(KJ2011Z292);安徽省自然科学基金(090416238)资助

摘　　要：针对智能监控中运动目标检测的问题,提出了基于Davinci-DM6467的高斯混合模型像素级的背景分割策略。对彩色图像建立高斯混合模型,根据场景中象素点的稳定性来调整模型参数的更新速率;通过和马氏阈值进行对比来判断是不是要更新背景模型;通过和背景阈值进行对比来判断哪几个模型是属于背景区域。经验证性实验测试,结果表明,高斯混合模型在运动检测中实时性好,对环境有较强的鲁棒性。In the light of movement target detection during intelligence monitoring, the author put forward Gaussian mixture model of pixel level background segmentation strategy based on Davinei-DM6467. Firstly establishing Gaussian mixture model for colorful images and then adjusting the updating velocity of model parameters according to the stability of each pixels in frames ; secondly comparing them with Markov threshold to judge whether to update the background model; finally comparing them with back-ground threshold to judge which several models belong to background region. Experimental results show that Gaussian mixture model in motion detection is possessed of good real-time and strong robustness to environment.

关 键 词：高斯混合模型 运动目标检测 马氏阈值 背景阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]