基于均值漂移聚类算法的背景重构

Background Reconstruction Based on the Mean Shift Clustering Algorithm

针对传统的基于像素点进行背景重构的算法中运算复杂、背景图像失真等缺点,提出了一种基于像素点均值漂移聚类算法的背景重构方法。首先通过变化率的判断省略一部分不存在运动对象的像素点,然后对其余大部分像素点需要通过典型样本点的均值漂移算法迭代运算,最后对结果进行聚类得到背景点值。仿真结果表明,该方法在复杂运动目标视频序列中能够快速、完整地重构背景。

To overcome disadvantages of traditional baekground reconstruction algorithms, such complex com- putation and distortional background, a background reconstruction algorithm was proposed based on pixel sequence mean shift clustering, first of all by the judgment of the rate of change a part of the pixels are omitted which the mo- tion object dose not exist, than the most of the pixels are calculed iteratively by typical sample point the mean shift algorithm, finally background point value are obtained from clustering results. The simulation results indicate the fast and efficient performance background reconstruction in complex moving target video sequence.