Mohamed Elrefaie
Experiments in Fluids · Springer Nature 2024

Real-Time and On-Site Aerodynamics using Stereoscopic PIV and Deep Optical Flow Learning

Mohamed Elrefaie, Steffen Hüttig, Mariia Gladkova, Timo Gericke, Daniel Cremers, Christian Breitsamter

Technical University of Munich · Volkswagen Group

Real-time
Velocity fields as they happen
Stereo PIV
3-component measurements
Deep optical flow
Learning replaces cross-correlation
The method

Taking the wind tunnel to the measurement, in real time

Particle image velocimetry is the gold standard for measuring flow fields — but classical cross-correlation processing keeps it offline and lab-bound. This work pairs stereoscopic PIV with deep optical-flow networks to deliver dense, three-component velocity fields in real time and on site, opening experimental aerodynamics to settings where conventional PIV processing is too slow.

On-site stereoscopic PIV: dense velocity fields reconstructed in real time with deep optical flow.
On-site stereoscopic PIV: dense velocity fields reconstructed in real time with deep optical flow.
Contributions

What the paper delivers

Learning-based PIV processing

Deep optical-flow models estimate dense particle displacement fields directly from image pairs, replacing iterative cross-correlation.

Real-time stereoscopic pipeline

End-to-end processing fast enough for live, on-site measurement of three-component velocity fields.

Validated against classical PIV

Accuracy benchmarked against state-of-the-art conventional processing, establishing trust in the learned estimates.

Reference

Citation

@article{elrefaie2024piv, title = {Real-time and on-site aerodynamics using stereoscopic PIV and deep optical flow learning}, author = {Elrefaie, Mohamed and H{\"u}ttig, Steffen and Gladkova, Mariia and Gericke, Timo and Cremers, Daniel and Breitsamter, Christian}, journal = {Experiments in Fluids}, volume = {65}, year = {2024}, publisher = {Springer Nature} }
Keep exploring

Related work