Can decode NOAA-satellite apt-data-images by parsing IQ-Wav files that were recorded by SDRSharp.
I might create a new repository later with automatic decoding. (and reference it here) Theoretically the signal can be improved if the model was trained at same location (with similar signal interferences). Additionaly no change of variables should be needed.
I should create something similar like my trained models on:
https://www.kaggle.com/crazyt/automatic-picture-transmission-decoding
(used example picture and static defined position)
https://www.kaggle.com/code/crazyt/masked-automatic-picture-transmission-decoding
(used static defined position)
https://www.kaggle.com/code/crazyt/search-masked-am-signal-apt-decoding
(takes damn long but finds a good position - at time of writing it is still a bit buggy finding a 100% correct position, but atleast creates an image)
But I would need to search the FM-signal in the band. So AM-based model cannot be used, sadly. I would need to create a similar IQ-wav-file first (so I'm 100% shure about the data of the original image). This should be possible with gnu-radio.
Problem is that I might need 2 models.
One model that finds the frequency of the signal in the Baseband and one that actually decodes the FM-signal.
Other resources (not by me) about NOAA:
https://www.cder.dz/download/Art7-1_1.pdf
https://www.noaasis.noaa.gov/NOAASIS/pubs/Users_Guide-Building_Receive_Stations_March_2009.pdf