Event-based imaging for visualization and measurement of turbulent flows

Abstract

Event-based vision (EBV), dynamic vision sensing (DVS) or neuromorphic imaging describe a rather new sub-field within computer vision, differing considerably from classical frame-based imaging (Gallego et al., 2022). Event cameras only record contrast changes ("events") within the scene, either going from dark to bright (positive event) or bright-to-dark (negative event). As the pixels "fire" independently an asynchronous stream of events results that consist of pixel coordinates, a time stamp and a binary contrast change signal. Static areas in the imaged scene provide no information; intensity data is essentially not available. At the same time, event cameras feature a very high dynamic range (>110 dB) and are considerably more sensitive than conventional CCD/CMOS cameras. In the context of particle imaging, narrow event streaks are produced in the space-time domain and can be processed to provide 3D-3C particle tracking velocimetry (PTV) data.The recently introduced event-based imaging velocimetry (EBIV) technique combines EBV and light sheet illumination to provide time-resolved, planar (2D-2C) velocity fields (Willert & Klinner 2022; Willert 2023). In this work we apply EBIV to obtain time-resolved velocity profiles of a turbulent boundary layer (TBL) in analogy to the profile-PIV technique (Willert, 2015). The latter has been used to simultaneously provide detailed velocity statistics and time-resolved data of turbulent flows (see eg. Willert et al. 2017). The field of view is generally illuminated by a high-speed pulsed laser that is collimated into a narrow light sheet. A second EBIV configuration captured the flow in the viscous sublayer of the TBL using a thin wall-parallel light sheet of <1 mm thickness and a set of three synchronized event-cameras. The ultimate aim of this setup is to estimate the unsteady wall shear stress field through triangulation of the recorded particle tracks which are addressed in this contribution. The dynamics of the near wall flow can already be visualized in the raw event data.