Classification and Feature Extraction of Lightning Electric Field Waveforms Based on Machine Learning

Abstract

Forecasting and warning of thunderstorms is very important to reduce the threat and damage of lightning to humans. The basis and prerequisite for forecasting and warning is the rapid identification and classification of data observed at multiple stations, extraction of waveform feature parameters and transmission them back to the central station. Currently, machine learning is a popular method and technique to achieve image recognition, classification and feature extraction. In this paper, based on the observed data of lightning electric field and machine learning, an image recognition model is constructed using convolutional neural network (CNN), and the recognition rate of the image is improved by stepwise optimization. The feature parameters of lightning are extracted based on OpenCV image processing techniques for subsequent real-time lightning localization, and can also be used to verify the classification results of lightning waveforms. The results show that the recognition rate of the final classification model can reach more than 90%, and the required waveform features can be extracted. This work has important application value and practical significance for the prediction and warning of lightning process observation.