基于改进光流法的雷达图像运动估计  

Motion Estimation for Radar Image Based on Improved Optical Flow Method

作　　者：王志斌[1] 肖艳姣[1] 吴涛[2] 

机构地区：[1]中国气象局武汉暴雨研究所,湖北武汉430205 [2]武汉中心气象台,湖北武汉430074

出　　处：《计算机技术与发展》2017年第12期170-175,共6页Computer Technology and Development

基　　金：国家自然科学基金资助项目(41275106);科技部行业专项公益性行业(气象)科研专项(GYHY201306008);湖北省雷电专项(FL-Z-201401)

摘　　要：介绍了一种改进的光流方法对天气雷达图像运动进行估计,计算天气雷达图像运动的光流变化,从而得到矢量场。用传统的Horn&Schunck方法中的全局平滑化算子估计时,其对数据的噪声敏感,且难以达到全场最优条件。通过加入改进后的高价平滑算子,可以减少靠向二级最小值的可能。另外,通过加入Lucas&Kanade方法计算局部光流得到小范围的光流场,在小的区域内光流场容易满足最优条件。适当运用平滑处理得到整场的矢量场,并以它作为改进Horn&Schunck方法的初猜场进行计算,且应用二维连续无辐散质量方程约束,同时对得到的风场进行风速以及风向的质量控制,得到比较完整的雷达运动矢量场。通过实验分析发现,改进的光流法优于气象上传统的交叉相关法,且优于单一使用某种光流方法的结果。An improved optical flow method is introduced to estimate the radar images motion and calculate the optical flow change of motion for radar image,obtaining the vector field. It is sensitive to data noise and difficult to meet the optimal condition when global smoothing operator from traditional Horn ＆ Schunck method is used to estimate. After adding the improved smoothing operator with high level, it can reduce the possibility for the secondary minimum. In addition, through Lucas ＆ Kanade method for calculation of the local optical flow, a small range of optical flow field is obtained, which can meet the optimal requirements. Smooth processing is applied to get the entire vector field appropriately, with it as fwst guess filed of improved Horn ＆ Schunck method for calculation. At the same time, through the two-dimensional continuous quality equation constraints without radiation,controlling the wind speed and direction of wind field acquired,the more complete motion vector field for radar is obtained. Experimental analysis shows that the improved optical flow method is better than traditional cross-correlation method and a single optical flow method.

关 键 词：雷达图像 运动估计 光流法 约束 改进 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]