利用特征向量聚类的流体运动矢量计算  

Calculation of Fluid Motion Vector Using Feature Vector Clustering

作　　者：全红艳[1] 

机构地区：[1]华东师范大学软件学院,上海200062

出　　处：《计算机辅助设计与图形学学报》2013年第2期221-228,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：上海市自然科学基金(11ZR1411100)

摘　　要：为了获得流体运动的二维矢量,利用视频流的两帧图像进行流体运动矢量的计算,提出一种基于特征向量聚类的流体运动矢量计算方法.首先利用主成分匹配的相关运算方法对流体运动矢量进行初始化;然后利用基于特征向量的聚类方法对初始化的运动矢量进行聚类分析,得到运动矢量的主要类别;最后利用最短距离的线性插值对主要类别中的运动矢量进行插值处理,得到密集的流体运动矢量的结果.实验结果表明,该方法具有方便、有效的特点,可用于流体运动场景重建与交互的后续研究.In order to obtain two-dimensional motion vector from two frames of video, a method of calculating fluid motion vector based on feature vector clustering is proposed in this study. First, fluid motion vector is initialized using correlation calculation of principal component matching. Then, it analyzes the initialized results with feature vector-based clustering to obtain the main categories of the motion vector. At last, it uses the linear interpolation method of minimum distance to process the motion vectors in the main categories and obtain the dense fluid motion vector. Experimental results show that the method has the characteristics of convenience and effectiveness. It can be used in the later study of reconstruction and interactive of fluid scene.

关 键 词：流体 运动矢量 最小距离聚类 线性插值 

分 类 号：TP391.71[自动化与计算机技术—计算机应用技术]