摘要
为了提高大密度梯度场的背景定向纹影(Background Oriented Schlieren,BOS)图像位移估计的精度和鲁棒性,提出了一种基于梯度恒定性假设的全局光流算法。首先,在传统光流算法的数据项(亮度恒定假设)中加入了梯度恒定假设;然后,采用改进的一阶div-curl正则化项作为空间平滑约束条件;最后,使用标准的“由粗到精”的多分辨率方案处理流场中的大位移。本方法对大密度梯度场的小波光学扭曲的合成BOS图像进行了测试,并与计算机视觉领域的经典光流算法的结果进行了比较。结果表明:均方根误差(Root Mean Square Error,RMSE)降低了0.131,结构相似性指数(Structure Similarity Index Measure,SSIM)提高了10.5%,在计算大密度梯度流场微弱的流动细节方面也更具有优势。
In order to improve the accuracy and robustness of background oriented schlieren(BOS)image displacement estimation in large density gradient field,a global optical flow algorithm based on gradient constancy assumption is proposed.Firstly,the gradient constant assumption is added to the data term(brightness constant assumption)of the traditional optical flow algorithm.Then,the improved first-order div-curl regularization term is used as the spatial smoothing constraint condition.Finally,the standard'coarse-to-fine'multi-resolution scheme is used to deal with the large displacement in the flow field.This method is used to test the synthetic BOS image of wavelet optical distortion of large density gradient field,and the results are compared with the classical optical flow algorithm in the field of computer vision.The results show that the root mean square error(RMSE)is reduced by 0.131,the structure similarity index measure(SSIM)is increased by 10.5%,and it has more advantages in calculating the weak flow details of large density gradient flow field.
作者
李虎
闫冠宇
朱海东
LI Hu;YAN Guanyu;ZHU Haidong(Guilin Institute of Information Technology,Guilin,Guangxi 541199,China;Guilin University of Electronic Technology,Guilin,Guangxi 541004,China;Guangxi Key Laboratory of Manufacturing System&Advanced Manufacturing Technology,Guilin,Guangxi 541004,China)
出处
《自动化与仪器仪表》
2025年第8期50-54,共5页
Automation & Instrumentation
关键词
背景定向纹影
梯度恒定假设
光流算法
大密度梯度场
background oriented schlieren
gradient constancy assumption
optical flow
large density-gradient field
作者简介
李虎(1998-),男,安徽宿州人,硕士,助教,研究方向为流体流量测量。
