中文摘要：

利用Horn—Schunck变分光流的通用模型，将描述符配准作为改进方法的支撑，结合经影像特征匹配所提取的同名特征，在光流条件中增加同名特征一致限制，提出一种基于特征匹配的变分光流方法。利用无人机视频影像对所提出的方法与原始变分光流算法进行了对比实验分析，实验结果表明该方法在目标位移过大时，能利用特征点信息解决光流连续性被打破的问题，准确获得大位移光流。

英文摘要：

Using the variational optical flow model introduced by Horn-Schunck, taking the descriptor matching as the support of the improved method and combining the descriptor correspondences from descriptor matching, a variational optical flow method was proposed. Some experiments of comparing and analyzing this new method with the traditional variational optical flow were carried out with UAV video sequence, and the results showed that it can solve the problem that the continuity of optical flow was broke when the object bad large dis- placement and get large displacement optical flow accurately.