摘要
利用计算机仿真粒子图像精确控制仿真粒子亚像素位移,讨论不同窗口下二次多项式曲面拟合法与基于梯度亚像素位移算法的匹配精度。针对粒子均匀移动的理想情况,发现二次多项式曲面拟合法的匹配精度随匹配窗口的增大而提高,而基于梯度亚像素位移算法的匹配精度几乎不随匹配窗口的增大而改变,两种算法匹配精度均在0.1个像素内。分析存在理想涡旋的粒子图像发现:二次多项式曲面拟合法的匹配精度约为0.2个像素,基于梯度亚像素位移算法的匹配精度约为0.05个像素。以二次多项式曲面拟合法为例,探讨了粒子大小和浓度对测量结果的影响,发现匹配窗口越大,最合适的粒子浓度范围就越大,这与不同匹配窗口包含的粒子图像灰度值信息量不同有关。在粒子大小方面,不同匹配窗口下,粒子直径为3个像素时误差均较小,表明PIV实际使用时,选择大小为3个像素的粒子为宜。将2种亚像素匹配算法应用到水槽实验获取的PIV影像分析中,均能得到较好的效果。
Two subpixel displacement matching methods of PIV,i.e.,the quadratic polynomial surface fitting method and subpixel displacement algorithm based on gray gradient,are compared in this paper.In the case where all of the particles move toward the same direction with the same displacement,the subpixel displacements and errors of the two methods are given under different windows,which shows that the displacement accuracy of the quadratic polynomial surface fitting method increases with the increasing of matching windows while the accuracy of subpixel displacement algorithm based on gray gradient shows little variations with the matching windows.The absolute mean displacement errors of the two methods are smaller than 0.1 pixel.For the case of idealized vortex flow,it is found that the displacement accuracy of the quadratic polynomial surface fitting method becomes within 0.2 pixels,and the accuracy of subpixel displacement algorithm based on gray gradient is about 0.05 pixels.Both particle sizes and concentrations influence the matching accuracy,results show that the larger the matching window,the larger the range of the suitable particle concentration,which is believed to be related to the valuable information of the particle image contained in different matching windows.Moreover,the particles with diameter of 3 pixels are found to be suitable for matching windows of 16 pixels×16 pixels,32 pixels×32 pixels and 64 pixels×64 pixels.The two subpixel matching methods are applied to analyze the actual PIV images obtained in wave tank experiments,both of them can achieve reasonable results.
作者
李超
蒋暑民
马洪余
戴德君
黄传江
LI Chao;JIANG Shu-min;MA Hong-yu;DAI De-jun;HUANG Chuan-jiang(First Institute of Oceanography,MNR,Qingdao 266061,China;Laboratory for Regional Oceanography and Numerical Modeling,Pilot National Laboratory for Marine Science and Technology,Qingdao 266237,China;Key Laboratory of Marine Science and Numerical Modeling,MNR,Qingdao 266061,China)
出处
《海洋科学进展》
CAS
CSCD
北大核心
2021年第1期9-19,共11页
Advances in Marine Science
基金
国家重点研发计划项目——“两洋一海”区域超高分辨率多圈层耦合短期数值预报系统研制(2017YFC1404000)
中央级公益性科研院所基本科研业务费专项资金——束星北青年学者人才项目(2018S03)
国家自然科学基金项目——卷波破碎致湍流混合的实验室实验研究(41376036)。
关键词
亚像素匹配精度
二次多项式曲面拟合法
基于梯度的亚像素位移算法
仿真粒子直径与浓度
subpixel displacement
quadratic polynomial surface fitting method
subpixel displacement algorithm based on gray gradient
particle diameter and concentration