<p>Event-based cameras are dynamic vision sensors that acquire asynchronous, pixel-wise data when local changes in light intensity exceed a user-defined threshold. Their high temporal resolution and data efficiency are particularly advantageous for applications with rapid movement of sparse/discrete signals such as in flow velocimetry. While previous studies have demonstrated the application of event-based particle image velocimetry (PIV) and particle tracking velocimetry (PTV), the impact of key hardware limitations, specifically latency and bandwidth, on the velocity dynamic range remains underexplored. This work first characterizes event cameras using single-pulse planar illumination of a water/alumina suspension in a cuvette, where effects of laser pulse energy, particle image density/size, and contrast threshold are systematically investigated. A camera performance map showed that non-retrievable information loss occurred row-wise for event rates above <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="348_2025_4127_Article_IEq1.gif" Format="GIF" Height="6" Rendition="HTML" Resolution="72" Type="Linedraw" Width="17" /> </InlineMediaObject> <EquationSource Format="TEX">$\sim$</EquationSource> <EquationSource Format="MATHML"><math> <mo>∼</mo> </math></EquationSource> </InlineEquation> 400 ev/<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="348_2025_4127_Article_IEq2.gif" Format="GIF" Height="13" Rendition="HTML" Resolution="72" Type="Linedraw" Width="12" /> </InlineMediaObject> <EquationSource Format="TEX">$\upmu$</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="normal">μ</mi> </math></EquationSource> </InlineEquation>s as the readout interface saturated due to camera bandwidth. The observed correlation between event rate and camera latency constrained the minimum time interval between successive laser pulses, therefore limiting the velocity that can be effectively measured with event-based PIV. A subsequent set of experiments performed in an air jet with double-pulse PIV produced exploitable vector fields for a velocity dynamic range of 0.2–1.8 m/s (0.003–0.03 px/<InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="348_2025_4127_Article_IEq2.gif" Format="GIF" Height="13" Rendition="HTML" Resolution="72" Type="Linedraw" Width="12" /> </InlineMediaObject> <EquationSource Format="TEX">$\upmu$</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="normal">μ</mi> </math></EquationSource> </InlineEquation>s) with particle image densities above 0.018 ppp and uncertainties <0.4 px. Results indicate that camera latency and bandwidth introduce a complex trade-off between the time interval between laser pulses (velocity dynamic range) and particle image density (spatial resolution) to ensure robust and exploitable velocity measurements. The proposed methodology can be readily applied to evaluate other event-based cameras and serve as a practical guideline to set up and optimize event-based PIV.</p>

An evaluation of event-based cameras for particle image velocimetry

期刊论文
发表日期：

2025年10月25日

查看全文

Luisa Maya,
Luming Fan,
Antoine Durocher,
Bruno Savard,
Patrizio Vena

摘要

Event-based cameras are dynamic vision sensors that acquire asynchronous, pixel-wise data when local changes in light intensity exceed a user-defined threshold. Their high temporal resolution and data efficiency are particularly advantageous for applications with rapid movement of sparse/discrete signals such as in flow velocimetry. While previous studies have demonstrated the application of event-based particle image velocimetry (PIV) and particle tracking velocimetry (PTV), the impact of key hardware limitations, specifically latency and bandwidth, on the velocity dynamic range remains underexplored. This work first characterizes event cameras using single-pulse planar illumination of a water/alumina suspension in a cuvette, where effects of laser pulse energy, particle image density/size, and contrast threshold are systematically investigated. A camera performance map showed that non-retrievable information loss occurred row-wise for event rates above $\sim$ $\sim$ 400 ev/ $\upmu$ $μ$ s as the readout interface saturated due to camera bandwidth. The observed correlation between event rate and camera latency constrained the minimum time interval between successive laser pulses, therefore limiting the velocity that can be effectively measured with event-based PIV. A subsequent set of experiments performed in an air jet with double-pulse PIV produced exploitable vector fields for a velocity dynamic range of 0.2–1.8 m/s (0.003–0.03 px/ $\upmu$ $μ$ s) with particle image densities above 0.018 ppp and uncertainties <0.4 px. Results indicate that camera latency and bandwidth introduce a complex trade-off between the time interval between laser pulses (velocity dynamic range) and particle image density (spatial resolution) to ensure robust and exploitable velocity measurements. The proposed methodology can be readily applied to evaluate other event-based cameras and serve as a practical guideline to set up and optimize event-based PIV.