摘要： 在机载气象雷达早期开发中，数值天气模型是一个非常重要的环节，而天气模型中风场的准确性直接影响整个模型的准确性。为解决数值天气模型中的连续图像运动分析问题，提出一种基于运动场Helmholtz分解的低维流体运动估计方法。通过少量涡流粒子和源粒子的演化，得到光流场的低维参数化表达，即非旋转量和螺旋量基函数的线性组合，其中基函数由格林核梯度构成，系数值和基函数参数通过最小化代价函数获得。实验结果表明，与传统光流法相比，该方法的计算速度快了近4倍，风场更为准确地反映了实际的天气状况，在风场反演中更为可靠。
关键词:
光流,
图像运动分析,
运动估计,
机载气象雷达,
数值天气,
风场重建
Abstract: In the early development of airborne meteorological radar, numerical weather is a very important part, while the accuracy of wind field directly affects the accuracy of the entire numerical weather model. Aiming at sloving the motion analysis of successive images in numerical weather model, low-dimensional fluid motion estimation method is used, and it is based on the Helmholtz decomposition of motion field. Through the deformation of a small number of vortex and source particals, it can get the low-dimensional parametric expression of optical flow field. The optical flow field consists of linear combinations of irrotational and solenoidal basis functions, which are based on Green kernel gradient. The coefficient values and the basis function parameters are obtained by minimization of a function. The experimental results show that compared with the traditional optical flow method, this method is nearly 4 times faster, and wind field more accurately reflects the actual weather conditions. The method is more reliable in wind field reconstruction.
Key words:
optical flow,
image motion analysis,
motion estimation,
airborne meteorological radar,
numerical weather,
wind field reconstruction
中图分类号: