基于光流法的流体运动预测技术适用性研究

Study on Applicability of Fluid Motion Prediction Technology Based on Optical Flow Method

由于粒子图像测速技术在测量精度和时空分辨率上难以同时保证,而基于光流法的流体运动估计方法可以获得具有较高分辨率的稠密速度矢量场。为了评价该方法在流体运动估计上的适用性,采用流动情况由简单到复杂的开源粒子图像、标量图像、纹影图像序列对3种经典光流算法进行实验,在定性和定量层面评估了各算法的表现。结果表明:传统HS光流算法适用于大部分流动情况,然而在灰度梯度变化不明显时表现不佳;基于随机湍流传输模型的变分光流算法表现更为稳定,在HS目标函数中添加加权非局部平滑项后更适用于捕捉涡轮叶片纹影图像的尾迹区域和估计尾迹涡的输运速度。

Unlike particle image velocimetry,which has difficulty in guaranteeing both measurement accuracy and spatio-temporal resolution,fluid motion estimation based on optical flow method can obtain dense velocity vector field with high resolution.To evaluate the applicability of this method in fluid motion estimation,three classical optical flow algorithms are tested by using open source particle image,scalar image and schlieren image sequences with flow conditions from simple to complex,and the performance of each algorithm is evaluated in quality and quantity.The results show that the traditional HS optical flow algorithm is suitable for most flow conditions,but performs poorly when the gray gradient changes are not obvious,the variational optical flow algorithm based on random turbulent flow model is more stable,and it is more suitable to capture the wake region of the turbine blade schlieren image and estimate the transport velocity of the wake vortex after adding the weighted nonlocal smoothing term to the HS objective function.