Global velocity measurement of fluorescent oil film based on deep learning optical flow method
-
摘要:
针对基于先验的传统光流法存在前提条件苛刻的问题,提出使用基于深度学习的光流法进行荧光油膜全局速度测量。采用数值仿真试验对基于先验的改进HS光流法和基于深度学习的FlowNet2光流法进行对比,结果显示:在不外加干扰时,改进HS光流法和FlowNet2光流法的平均端点误差分别为0.458 7像素/s和0.381 7像素/s;在亮度变化、噪声干扰或不同的演化时间下,FlowNet2光流法的平均端点误差均明显低于改进HS光流法,平均端点误差差值最大可达5.19像素/s;风洞试验进一步证明,FlowNet2光流法能够获得正确、清晰、定量的荧光油膜全局速度场,较改进HS光流法鲁棒性更高,对风洞工程应用具有一定的参考价值。
Abstract:In order to solve the problems of the traditional optical flow method based on priori,such as harsh preconditions,an optical flow method based on deep learning was proposed to measure the global velocity of fluorescent oil film.The numerical simulation experiments were used to compare the improved HS optical flow method based on prior with FlowNet2 optical flow method based on deep learning.The results showed that the average endpoint errors of improved HS optical flow method and FlowNet2 optical flow method were 0.458 7 pixel/s and 0.381 7 pixel/s without external interference,respectively;the average endpoint error of FlowNet2 optical flow method was significantly lower than that of HS optical flow method under the conditions of brightness change,noise disturbance or different evolution times,and the maximum difference of average endpoint errors could reach 5.19 pixel/s.The experimental results in the wind tunnel further prove that the FlowNet2 optical flow method can obtain the correct,clear and quantitative global velocity fields of fluorescent oil film.With stronger robustness than the improved HS optical flow method,this method has certain reference value for wind tunnel engineering application.
-
Key words:
- optical flow method /
- deep learning /
- fluorescent oil film /
- global velocity /
- robustness
-
表 1 两种数据集特征
Table 1. Features of two datasets
特征 数据集名称 Chairs Things3D 训练集帧数 22 232 21 818 测试集帧数 640 4 248 场景数 964 2 247 表 2 不同亮度变化下的平均端点误差
Table 2. Average endpoint error with different brightness changes
亮度变化/% 平均端点误差/(像素/s) 改进HS光流法 FlowNet2光流法 -10-8-6-4-20+2+4+6+8+10 0.557 00.555 90.557 40.556 30.559 40.458 70.579 30.620 90.670 90.752 80.840 0 0.495 60.477 50.459 50.452 70.447 90.381 70.450 30.452 30.458 40.461 90.464 0 表 3 不同噪声下的平均端点误差
Table 3. Average endpoint error with different noises
噪声类型 平均端点误差(像素/s) 改进HS光流法 FlowNet2光流法 无噪声 0.458 7 0.381 7 高斯噪声 0.775 4 0.457 9 椒盐噪声 0.859 8 0.460 4 混合噪声 0.855 5 0.456 6 表 4 不同演化时间下的平均端点误差
Table 4. Average endpoint error with different evolution times
演化时间/s 平均端点误差/(像素/s) 改进HS光流法 FlowNet2光流法 0.020.040.060.080.100.120.140.160.180.20 0.458 60.564 40.718 61.426 12.495 63.709 54.882 76.091 17.419 08.632 7 0.381 70.436 00.527 70.650 50.858 41.203 31.674 12.197 92.760 33.442 7 -
[1] 邢宇,罗东明,余雄庆.超临界层流翼型优化设计策略[J].北京航空航天大学学报,2017,43(8):1616⁃1624.XING Yu,LUO Dongming,YU Xiongqing.Optimization strategy of supercritical laminar flow airfoil design[J].Journal of Beijing University of Aeronautics and Astronautics,2017,43(8):1616⁃1624.(in Chinese) [2] 朱自强,吴宗成,丁举春.层流流动控制技术及应用[J].航空学报,2011,32(5):765⁃784.ZHU Ziqiang,WU Zongcheng,DING Juchun.Laminar flow control technology and application[J].Acta Aeronautica et Astronautica Sinica,2011,32(5):765⁃784.(in Chinese) [3] 王帅,何国强,秦飞,等.超声速内流道摩擦阻力分析及减阻技术研究[J].航空动力学报,2019,34(4):908⁃919.WANG Shuai,HE Guoqiang,QIN Fei,et al.Research on skin⁃friction drag and drag reduction technics in a supersonic inner flow path[J].Journal of Aerospace Power,2019,34(4):908⁃919.(in Chinese) [4] 黄湛,王宏伟,魏连风,等.基于荧光油膜的全局表面摩阻测量技术研究[J].空气动力学学报,2016,34(3):373⁃378.HUANG Zhan,WANG Hongwei,WEI Lianfeng,et al.Research of global skin friction measurement based on fluorescent oil film[J].Acta Aerodynamica Sinica,2016,34(3):373⁃378.(in Chinese) [5] NAUGHTON J W,SHEPLAK M.Modern developments in shear⁃stress measurement[J].Progress in Aerospace Sciences,2002,38(6/7):515⁃570. [6] 林其勋,郭捷,李亚平,等.用热线风速仪研究轴流压气机的非定常流场及端壁附面层[J].航空动力学报,1989,4(2):141⁃144.LIN Qixun,GUO Jie,LI Yaping,et al.Investigation on unsteady flow field and endwall boundary layer in axial flow compressor with hot wire anemometer[J].Journal of Aerospace Power,1989,4(2):141⁃144.(in Chinese) [7] 徐大川,龙彦志,李玉栋.基于激光多普勒原理的极低风速测量实验[J].航空动力学报,2020,35(6):1228⁃1237.XU Dachuan,LONG Yanzhi,LI Yudong.Experiment of extremely low wind velocity measurement based on laser Doppler[J].Journal of Aerospace Power,2020,35(6):1228⁃1237.(in Chinese) [8] LIU T,SULLIVAN J P.Luminescent oil⁃film skin⁃friction meter[J].AIAA Journal,2015,36(8):1460⁃1465. [9] 钱泓江 董秀成 张征宇 基于Hankel阵的荧光油膜灰度与厚度预测模型改进 2021 36 10 2061 2071 Improved prediction model of gray and thickness of fluorescent oil film based on Hankel matrix 2021 36 10 1228 1237 (in Chinese) [10] QIAN H,DONG X C,ZHANG Z Y.Research on stability of gray value of excited⁃state fluorescent oil film based on variable light vector angle[J].Mathematical Problems in Engineering,2021:8632992.1⁃8632992.12. [11] 钱泓江,董秀成,徐椰烃,等.基于神经网络的荧光油膜厚度与灰度研究[J].南京航空航天大学学报,2021,53(3):470⁃476.QIAN Hongjiang,DONG Xiucheng,XU Yeting,et al.Fluorescence oil film thickness and gray level identification based on neural network[J].Journal of Nanjing University of Aeronautics and Astronautics,2021,53(3):470⁃476.(in Chinese) [12] ZHAI M,XIANG X,LV N,et al.Optical flow and scene flow estimation:a survey[J].Pattern Recognition,2021,114(3):107861.1⁃107861.15. [13] TU Z,XIE W,ZHANG D,et al.A survey of variational and CNN⁃based optical flow techniques[J].Signal Processing Image Communication,2019,72:9⁃24. [14] HORN B,SCHUNCK B G.Determining optical flow[J].Artificial Intelligence,1981,17(1):185⁃203. [15] SUN D,ROTH S,Black M J.Secrets of optical flow estimation and their principles[C]∥Proceeding of IEEE Conference on Computer Vision and Pattern Recognition.San Francisco,US:IEEE,2010:2432⁃2439. [16] 邵绪强,杨艳,刘艺林.流体运动估计光流算法研究综述[J].中国图象图形学报,2021,26(2):355⁃367.SHAO Xuqiang,YANG Yan,LIU Yilin.Review of optical flow algorithm in fluid motion estimation[J].Journal of Image and Graphic,2021,26(2):355⁃367.(in Chinese) [17] FISCHER P,DOSOVITSKIY A,ILG E,et al.FlowNet:learning optical flow with convolutional networks[C]∥Proceeding of IEEE International Conference on Computer Vision.Santiago,Chile:IEEE,2015:2758⁃2766. [18] ILG E,MAYER N,SAIKIA T,et al.FlowNet 2.0:evolution of optical flow estimation with deep networks[C]∥Proceeding of 2017 IEEE Conference on Computer Vision and Pattern Recognition.Honolulu,US:IEEE,2017:1647⁃1655. [19] HEITZ D,MÉMIN E,SCHNÖRR C.Variational fluid flow measurements from image sequences:synopsis and perspectives[J].Experiments in Fluids,2010,48(3):369⁃393. [20] BROX T,BRUHN A,PAPENBERG N,et al.High accuracy optical flow estimation based on a theory for warping[C]∥Proceeding of European Conference on Computer Vision.Berlin,German:Springer,2004:25⁃36. [21] 李鹏.全局表面摩擦应力直接测量技术研究[D].南京:南京航空航天大学,2012.LI Peng.Studies of direct measurement techniques about global skin friction[D].Nanjing:Nanjing University of Aeronautics and Astronautics,2012.(in Chinese) [22] 邹易峰,张征宇,王学渊,等.风洞试验模型表面的荧光油膜路径运动速度测量[J].航空学报,2019,40(6):92⁃105.ZOU Yifeng,ZHANG Zhengyu,WANG Xueyuan,et al.Velocity measurement of fluorescent oil film path movement on wind tunnel testing model surface[J].Acta Aeronautica et Astronautica Sinica,2019,40(6):92⁃105.(in Chinese) [23] WANG Bo,CAI Zemin,SHEN Lixin,et al.An analysis of physics⁃based optical flow[J].Journal of Computational and Applied Mathematics,2015,276:62⁃80. [24] 陈吉明,吕金磊,廖达雄,等.连续式跨声速风洞试验段降噪技术[J].航空动力学报,2021,36(8):1712⁃1719.CHEN Jiming,Jinlei LÜ,LIAO Daxiong,et al.Noise reduction technology in test section of continuous transonic wind tunnel[J].Journal of Aerospace Power,2021,36(8):1712⁃1719.(in Chinese) [25] HU Y,RUI S,LI Y.Efficient coarse⁃to⁃fine patch match for large displacement optical flow[C]∥Proceeding of IEEE Conference on Computer Vision and Pattern Recognition.Las Vegas,US:IEEE,2016:5704⁃5712. [26] WEINZAEPFEL P,REVAUD J,HARCHAOUI Z,et al.DeepFlow:large displacement optical flow with deep matching[C]∥Proceeding of IEEE International Conference on Computer Vision.Sydney,Australia:IEEE,2013:2523⁃2541. [27] COSTANTINI M,LEE T,NONOMURA T,et al.Feasibility of skin⁃friction field measurements in a transonic wind tunnel using a global luminescent oil film[J].Experiments in Fluids,2021,62(1):1⁃34.