基于高斯混合模型的视频运动对象自动分割算法  

Video Moving Object Segmentation Based on Gaussian Mixture Model

作　　者：李新仕[1] 王天江[1] 刘芳[1] 

机构地区：[1]华中科技大学计算机科学与技术学院智能与分布计算实验室,武汉430074

出　　处：《计算机科学》2009年第1期205-207,共3页Computer Science

基　　金：国家高技术研究发展(863)计划(项目编号:2007AA01Z161)的支持

摘　　要：提出的算法首先采用高斯混合模型依据空间属性对当前帧进行聚类分割,可以克服一般聚类算法对数据集中的噪声无法建模以及聚类数目难以确定的问题。然后依据时序属性,分割出当前帧的运动对象的初步轮廓区域。最后将初步轮廓区域和聚类分割的区域进行匹配,提取出视频运动对象。通过实验验证,算法具有较好的准确性和抗干扰性,在运动微小的情况下也能取得比较好的效果。An automatic and accurate video moving object segmentation algorithm was proposed which is based on the spatio temporal property segmentation. The Gaussian Mixture Model was employed to cluster the current frame into a set of homogeneous regions based on the spacial property. Comparing with the general cluster algorithms, it could model the noise in the data set and estimate the optimal number of clusters more easily. It extracted the preliminary contour region of the moving object in the current frame based on the temporal property. The video moving object was extracted accurately by matching the homogeneous regions with the preliminary contour region. The experiment results demon- strate that the proposed algorithm is not only accurate and robust to noise, but also effective even if the movement is little.

关 键 词：视频运动对象 高斯混合模型 最短描述长度 运动估计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] O241.82[自动化与计算机技术—计算机科学与技术]