Event-based imaging velocimetry: A new approach to flow diagnostics based on particle imaging

Abstract

Event-based vision (EBV), also referred to as dynamic vision sensing (DVS), is a rather recent emerging, bio-inspired technology in the field of computer vision that has already found numerous applications in areas such as autonomous navigation, high-speed object counting, star tracking, 3D sensing or vibration monitoring. Unlike the framing sensor in a conventional camera, the detector of an EBV camera only reacts to contrast changes producing "on" or "off" events at the pixel level with microsecond resolution. With each pixel operating individually, the output data stream is asynchronous, containing pixel coordinates, event time and polarity of the contrast change. In the context of imaging small particles in fluid flows, the event data stream essentially provides continuous real-time particle tracking data as already demonstrated by a few other research groups. The focus of this presentation is the assessment of EBV in the context of providing dense fields of time-resolved velocity data and includes demonstrations on simple water and air flows. Several (off-line) processing algorithms for extracting flow field data from the asynchronous data stream are explored. Measurements at equivalent frame rates in the 2-5 kHz-range are feasible. The currently available hardware does however impose some limitations, most notably, when exceeding the available camera bandwidth while imaging densely seeded flows. This and other unique aspects of the new imaging approach will be addressed in the presentation along with possible mitigation strategies. Beyond the more quantitative applications on simple water and air flows explored here, EBV offers considerable potential as a tool for flow visualization and should be readily useable for educational purposes, in particular, in combination with eye-safe light sources such as LED illumination.