结合三帧差分与改进混合高斯模型的运动目标  

Combining with the Three-Frame Difference and Improved Gaussian Mixture Model of Moving Object Detection Method

作　　者：廖武[1] 赵新民[1] 钱盛友[1] 聂义[1] 

机构地区：[1]湖南师范大学物理与信息科学学院,长沙410081

出　　处：《电脑与信息技术》2016年第1期14-17,60,共5页Computer and Information Technology

基　　金：国家自然科学基金(项目编号:11474090);湖南省研究生科研创新项目(项目编号:CX2014B222);湖南省教育厅基金项目(项目编号:12C0204)

摘　　要：针对混合高斯模型对于噪声与光照变化检测效果不佳的问题,文章提出了结合三帧差分与改进型混合高斯模型的运动目标检测方法。该方法先通过三帧差分快速获取一副背景图像,然后将该背景图像按一定的比例更新到混合高斯模型主背景分布中,再按照改进的混合高斯模型进行背景提取,最后得到前景图像。实验结果表明,利用改进的混合高斯模型算法,提高了算法检测的准确度,并且结合三帧差分法能有效解决噪声与光照变化问题,提高了算法的鲁棒性。In order to deal with the problem which Gaussian mixture model for noise and light change detection effect is not good,a method for moving object detection that combined three-frame difference and improved Gaussian mixture model is proposed.Firsdy,a background image is quickly obtained by three-frame difference,then,update the background image in a certain proportion to the main background distribution of the Gaussian mixture model,then,extract a new background image according to the improved Gaussian mixture model,finally,get the foreground image.The simulation results show that using the improved Gaussian mixture model can improve the accuracy of moving object detection, and combine three-frame difference method can effectively solve the problem of noise and illumination changes and improve the robustness of the algorithm.

关 键 词：运动目标检测 三帧差分 混合高斯模型 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]