Observation and Analysis of a Multiple Lightning Strike Based on Dynamic Vision

To capture lightning more conveniently and efficiently, this paper employs dynamic vision sensor technology to collect lightning data. Compared to high-speed cameras, dynamic vision sensors have the advantages of smaller data volume and lower cost; compared to conventional cameras used in power grids, dynamic vision sensors offer higher temporal resolution. This study employs dynamic vision technology to observe and analyze lightning strikes. This article develops a set of event-based image reconstruction and clustering algorithms to analyze a multiple cloud-to-ground lightning strike that occurred on May 6, 2024, restoring the lightning process and extracting the lightning channel. Firstly, the event data is reconstructed into lightning images, revealing that this cloud-to-ground lightning event includes one leader discharge and three return strokes, with an average time interval of 60ms between each discharge. Additionally, the K-Means clustering algorithm is employed to extract the lightning channel portion from the pixel image. This research utilizes dynamic vision sensor technology to study the development and channel extraction of multiple lightning strikes, providing a new method and effective approach that can serve as a powerful complement to traditional lightning observation techniques.