Research on real-time display of flow field based on optical flow algorithm

In this paper, Lucas-Kanade (LK) optical flow algorithm with high efficiency and Liu-Shen (LS) optical flow algorithm with strong physical properties are combined to achieve high precision real-time display of the flow field. The efficiency, accuracy and robustness of the algorithm are tested by the synthetic particle image test and the towing flume device, which is compared with the cross-correlation algorithm. In the test of synthetic image data set, LK-I algorithm has the shortest computing time, only 15ms. LK-S&LS algorithm has the highest computational accuracy and robustness, and its root-mean-square error and angular errors are 0.073 and 1.56 respectively with the strongest anti-noise ability. Cross-correlation algorithm takes 1500ms, and its root-mean-square error and angular error are 0.128 and 3.13 respectively, so it is obviously inferior to the other two optical flow algorithms. In the experimental test based on the towing flume, both LK and LK&LS coupling algorithms can quickly display the velocity field and vorticity field in the near-wake region. The results show that LK&LS algorithm has the least flow field noisy points, while cross-correlation algorithm has the most noisy points. This research can also provide some algorithm support for the realization of blunt body drag and noise reduction in the following real-time control problems.