Automatic Characterization of Mid-latitude Multiple Ionospheric Plasma Structures from All-sky Airglow Images using Deep Learning Technique

TL;DR

This paper presents a fully automated deep learning-based method for detecting and characterizing multiple ionospheric plasma structures from all-sky airglow images, enabling efficient analysis of large datasets in mid-latitude regions.

Contribution

It introduces a novel automated pipeline combining YOLOv8 and BoT-SORT for localizing and tracking plasma structures, with a quality control step to ensure reliable parameter estimation.

Findings

The automated method accurately estimates propagation parameters of plasma structures.

The pipeline outperforms semi-automatic approaches in efficiency and reliability.

It effectively handles large datasets for mid-latitude ionospheric studies.

Abstract

The F-region ionospheric plasma structures are propagating high and or low electron density regions in the Earth ionosphere. These plasma structures can be observed using ground based all-sky airglow imagers which can capture faint airglow emissions originating from the F-region of ionosphere. This study introduces a novel automatic method for determining the propagation parameters (horizontal velocity and orientation) of these multiple ionospheric plasma structures observed in O(1D) 630.0 nm all-sky airglow images from Hanle, India located in the mid-latitude region. We have used a deep learning-based segmentation model called YOLOv8 (You Only Look Once) to localize and BoT-SORT tracker to track individual mid-latitude ionospheric plasma structures. Three different automatic algorithms are used to characterize the observed plasma structures utilizing the segmented outputs from the YOLO model. Finally, an additional quality control step is introduced that filters the results from the three automatic algorithms and generates a flag to retain the most reliable estimate. The results of the proposed fully automated pipeline are systematically compared with a previously developed semi-automatic approach to assess the estimation efficacy. The automatic technique developed in this study is particularly valuable for all-sky airglow imaging systems having large datasets, where manual intervention or semi-automatic analysis is impractical.