流体运动图像的物理光流计算方法研究

项目名称： 流体运动图像的物理光流计算方法研究

项目编号： No.61203249

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 自动化学科

项目作者： 蔡泽民

作者单位： 汕头大学

项目金额： 24万元

中文摘要： 基于光流的流体运动图像计算与传统PIV方法相比，能充分利用已有的各种流体可视化技术并显著提高流体运动矢量场的精度。然而由于光流只是对图像亮度模式运动的反映，简单地利用图像的光流作为真实运动的描述存在不足，有时甚至不能反映流体真实运动。此外，图像光流缺少物理含义，严重影响光流方法在流体运动图像计算与分析上的应用。为此，课题主要研究一般流体运动图像的物理光流计算方法：1）基于透射投影变换，从流体力学基本方程出发，推导流体运动的物理光流模型，为流体运动矢量场的精确计算奠定基础；2）针对图像域上的投影运动方程，结合流体运动的初始与边界条件信息设计合理的正则化约束条件，以保证对流体运动图像物理光流的正确求解，并转化为流体运动矢量；3）增强与复原流体运动图像，进一步提高流体运动矢量场的精度。课题的研究将增强光流方法在流体力学中的适用性，并为运动图像分析这一计算机视觉重要领域的研究和应用提供理论支撑。

中文关键词： 物理光流方法；正则化；运动分析；滤波；

英文摘要： Comparing with the traditional particle image velocimetry(PIV) method, fluid velocity determination based on fluid motion images using optical flow can improve the precision of velocity vector field significantly. Furthermore, all kinds of fluid visualization technologies will come in handy, but not only the PIV. However, the brightness constraint equation of optical flow is not derived from any physical principle for a physical process, and therefore it does not hold exactly from a physical point of view. More importantly, the optical flow in the brightness constraint equation does not have a clearly defined physical meaning, and as a result it cannot be generally and directly used for quantitative measurements of physical quantities from images for most scientific and engineering problems. All of these slow down the researches on optical flow and its applications on computation and analysis of fluid flow. For this reason, this project is to research on a generic model of optical flow for fluid motion images, which is named physics-based optical flow. The research contents will focus on the following three parts. Firstly, the mathematical model of physics-based optical flow will be constructed from the basic equations of fluid mechanics, using the perspective projection transformation. Secondly, constrained con

英文关键词： Physics-based optical flow；Regularization；Motion analysis；Filtering based；