.png){kind=icon}
.png){kind=icon}
Return
IncResUnet: A Model for Automatic Ionospheric Plasma Bubble Detection
.png){kind=icon}
OA
AI
DOI：10.34133/space.0492.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Plasma bubbles are important ionospheric phenomena in the equatorial region that can cause scintillation interference in high-frequency communication. Traditional methods for detecting plasma bubbles rely on measuring the deviation of ion density over a period. However, practical applications have revealed instances of missed detections and false alarms. To address this issue, we combine the detection results of the classical method with plasma density features collected from the Formosat-1 to establish a dataset of plasma bubbles and define new identification criteria. Moreover, we introduce innovative elements into the U-Net model, creating an automated detection model called IncResUnet. Experimental results demonstrate that IncResUnet achieves an F1-score of 0.91, a recall score of 0.95, and a precision score of 0.87, outperforming traditional detection methods. Furthermore, we successfully apply the IncResUnet model, trained on Formosat-1 data, to C/NOFS data, achieving excellent detection performance and demonstrating its robust generalization capability.
Journal
S
IF:
6.8
Papers:
255
Citations:
797