现代电子技术2016，Vol.39Issue(11)：63-69,7.DOI:10.16652/j.issn.1004-373x.2016.11.016
复杂场景下基于动态纹理的运动分割和背景估计方法
Motion segmentation and background estimation method based on dynamic texture in complex scenes
摘要
Abstract
Since the available motion segmentation and background estimation method can′t segment the static object,and is unsuitable for complex dynamic scenes,a foreground⁃background hybrid model(FBM)based on dynamic texture(DT)is pro⁃posed to realize the union expression of background and foreground in dynamic scenes. The FBM is composed of a set DT compo⁃nents about location and a set global DT components. The former is used to simulate the local background motion,and the latter is used to simulate the persistent foreground motion. The EM algorithm for learning FBM parameters and variation approximation strategy are proposed,which can realize the foreground motion segmentation in various moving complex scenes and detect the static object while FBM needn′t threshold artificial selection and individually training video. The simulation experiment results show that,in comparison with the latest motion segmentation and background estimation method,the proposed method can sig⁃nificantly improve the accuracy of background estimation and motion segmentation.关键词
运动分割/背景估计/动态纹理/EM算法/阈值/分割精度Key words
motion segmentation/background estimation/dynamic texture/EM algorithm/threshold/segmentation accuracy分类
信息技术与安全科学引用本文复制引用
夏蕾,罗佳..复杂场景下基于动态纹理的运动分割和背景估计方法[J].现代电子技术,2016,39(11):63-69,7.基金项目
国家自然科学基金 （）
