A POD-based spatial resolution enhancement method for real-time event-based imaging velocimetry

Abstract

Event-Based Imaging Velocimetry (EBIV) introduces a novel paradigm in flow diagnostics by replacing conventional frame-based image acquisition with asynchronous, high-frequency event data, enabling reduced latency and lower data rates for real-time applications. We propose a real-time method to estimate high-resolution (HR) flow fields from low-resolution (LR) EBIV data, based on POD-driven spatial refinement. In the offline phase, EBIV recordings are processed using both fast single-pass (LR) and accurate multi-frame, multi-pass (HR) cross-correlation. POD is applied to extract spatial modes and temporal coefficients. Online, new LR measurements are projected onto downsampled HR modes to estimate coefficients, which are then used to reconstruct the HR field. A Kalman filter refines the coefficients using a dynamic model learned from training data, enhancing temporal consistency and robustness. The methodology has been tested on two flow configurations: a turbulent channel flow in air at 5 kHz and a submerged water jet in water at 200 Hz. Results demonstrate the method's capability to deliver fast and accurate high-resolution estimations, making it suitable for real-time flow analysis in complex unsteady environments.