结合水平集与光流的序列影像运动分割技术研究

摘要

A motion segmentation method combining variational level set segmentation with variational optical flow was presented in this paper. The energy function of level set and the energy function of optical flow were integrated to form a new motion segmentation energy function, and the variational method was used to optimize the energy function. The experiments of analyzing the motion segmentation methods were carried out with UAV image se-quences.The results showed that this method not only accurately estimated the motion vectors, but also accurately segmented the target areas when the motion displacement of target areas was overlarge.%提出一种结合变分水平集分割与变分光流的运动分割方法。该方法将光流能量函数与水平集分割能量函数进行整合形成新的运动分割能量函数，并采用变分方法进行优化。采用无人机序列影像对该运动分割方法进行实验分析，实验结果表明该方法在目标运动位移过大时，既能准确地估计出运动矢量，又能精确地分割出目标区域。