Segmentation and Edge Detection for Ionogram Automatic Scaling

Ionograms, which are captured by ionosondes, include crucial data about the ionosphere. The first step in analyzing ionospheric weather is scaling ionograms. Due to the current rapid expansion of data collection, it is impossible for human experts to manually scale a sizable number of ionograms in a timely manner. In this study, a three-stage method based on semantic segmentation networks, an edge detection module, and a feature fusing module is suggested to obtain precise ionospheric parameters of the E, F1, and F2 layers. The segmentation networks are trained using 3448 ionograms from the Chinese Academy of Sciences Digital Ionosonde situated in Hainan, Huailai, and Wuhan. The test results over 863 images indicate that 99.1% of the critical frequency foF2 autoscaled values have an error of within 0.2 MHz, and 98.7
^%
of the minimum virtual height h'F2 autoscaled values have an error of within 10 km, demonstrating that our technique performs of almost as well as that of human specialists. Thus, the investigation into ionospheric physics could benefit greatly from our work.