Automated Large‐Scale Extraction of Whistlers Using Mask‐Scoring Regional Convolutional Neural Network

Abstract

Extremely and very low frequency (ELF/VLF, 3 Hz-30 kHz) radio signals play an important role in navy communications, ionospheric remote sensing, radiation belt dynamics, and several related geophysical applications (

Monitoring Space Weather: Using Automated, Accurate Neural Network Based Whistler Segmentation for Whistler Inversion

“…Whistler propagation 𝐴𝐴 𝐴𝐴 shells and their electron densities are coded in the whistler shape, elucidating physical parameters which cannot be extracted by detection alone. Harid et al (2021) segmented whistler traces on spectrogram images using an MSRCNN, although the result was used only for counting them, not for actually inverting the whistlers and estimating their physical parameters. The main added value of our convolutional neural network model is the automated inversion of whistler traces, thus providing electron density estimates over a range of 𝐴𝐴 𝐴𝐴 shells.…”

“…Recently (Harid et al., 2021) developed a mask‐scoring regional convolutional neural network (MSRCNN) for the automated extraction of whistlers, but the extracted f ‐ t pairs were not used to obtain the equatorial electron densities through whistler inversion. In this study, we present a method which is similarly based on machine learning, leveraging the PointRend (Kirillov et al., 2020) architecture.…”

Statistical Analysis of Electromagnetic Ion Cyclotron Rising‐Tone Emissions Based on Deep Learning

“…In the past two decades, deep learning has also been used in the field of space physics due to the improvement of computing performance with Graphics Processing Units (GPUs), progress in the learning methods, and the support of the vast data set from in situ spacecraft measurements and ground-based observations. Deep learning is playing important roles in many space physics studies, such as the classification of aurora images (Clausen & Nickisch, 2018;Kvammen et al, 2020), the construction of magnetospheric plasma density models (Bortnik et al, 2016(Bortnik et al, , 2018, the detection of interplanetary coronal mass ejections (Nguyen et al, 2019), the classification of magnetospheric plasma regions (Breuillard et al, 2020), and the identification of plasma waves such as whistlers (Harid et al, 2021), chorus and hiss (Bortnik et al, 2018), and EMIC waves (Medeiros et al, 2020).…”

Automatic Detection of Atmospherics and Tweek Atmospherics in Radio Spectrograms Based on a Deep Learning Approach

“…A large amount of data is needed for statistical studies to confirm the observed trends. Machine learning techniques can significantly simplify the detection and analysis of the sferics and tweeks or whistlers (Harid et al, 2021) recorded in the form of frequency-time spectrograms. The research presented below focuses on an automatic method for detection and localization of specific signals within frequency-time spectrograms based on convolutional neural networks (LeCun et al, 2015).…”