RapidPIV: Full Flow-Field kHz PIV for Real-Time Display and Control
ORAL
Abstract
We present a novel architecture for accelerating particle image velocimetry (PIV) calculations. An optical flow hardware accelerator calculates initial flow estimates, with cross-correlation providing corrections in post-processing. The result is RapidPIV: a software program for real-time PIV on Nvidia Turing-gen (or newer) GPUs. Dense vector fields and vorticity are displayed in real time when connected to a compatible camera. We achieve 1,150 frames-per-second on 1 megapixel images, with latency of 1ms. Accuracy, repeatability, and robustness are tested with physical experiments involving an accelerated flat plate and a cylinder wake. RapidPIV's accuracy, precision and ability to handle high displacement, velocity gradients, out-of-plane motion, and low seeding density compare favorably with the trusted multi-grid correlation software, while performing thousands of times faster than CPU-based software.
*Support was provided in part by the Center for Autonomous Systems and Technologies (CAST) at Caltech. This material is based upon work supported in part by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-1745301.
–
Publication: https://arxiv.org/abs/2504.17987
Manuscript submitted to Measurement Science and Technology
Presenters
-
Samuel H Foxman
- Caltech