Dynamic vision sensing for impulse current inversion: a novel approach leveraging event-based optical signal analysis in gas discharge experiments

Abstract

Lightning is a brief discharge event that occurs in the atmosphere, characterized by tremendous energy and electromagnetic effects. Lightning current is a critical parameter holding great importance for both lightning physics research and protection design. However, direct field measurements of lightning current face considerable challenges. Therefore, inversion based on optical signals has become a primary method for obtaining lightning current information: by measuring the optical radiation from the discharge channel and establishing a luminosity-current relationship, the transient current process can be estimated indirectly. Existing optical monitoring methods encounter issues such as large data volumes, high power consumption, limited interference resistance, and restricted monitoring coverage. Dynamic Vision Sensing (DVS), as a neuromorphic imaging technology, offers advantages such as ultra-high dynamic range, low data redundancy, and high temporal resolution, demonstrating potential for monitoring fast transient events.

This study introduces DVS technology to the impulse-current discharge experiments. An impulse current generator was used to produce decaying oscillatory impulse currents with slow-rising fronts, applied to a 5 mm point-to-point discharge gap to trigger gas discharge. Simultaneously, dynamic vision data and discharge channel current were measured to investigate the correlation between impulse currents and event data. The DVS generates ON/OFF events based on changes in light intensity. The results reveal that the cumulative value of ON events exhibits a positive correlation with the peak current, with a Pearson correlation coefficient greater than 0.92. Conversely, the cumulative values of OFF events and ALL events show a negative correlation with the peak current, with Pearson correlation coefficients greater than 0.95, and quadratic correlation coefficients greater than 0.97 in both cases. More importantly, this study derives a mathematical relationship between impulse currents and event counts, achieving preliminary inversion of impulse currents. This work marks the first application of dynamic vision sensing technology to investigate the inversion of large impulse currents via event data, providing a novel approach for monitoring impulse currents and even lightning currents.