Editorial Office:
Management:
R. S. Oyarzabal
Technical Support:
D. H. Diaz
M. A. Gomez
W. Abrahão
G. Oliveira
Publisher by Knobook Pub
doi: 10.6062/jcis.2019.10.03.0165
(Free PDF)A.L.G. Pereira, C.G. Giménez de Castro and J.F. Valle Silva
The Solar Submillimeter Telescope (SST) observes simultaneously and independently with a multibeam focal array at 212 and 405 GHz. Since 1999, the SST daily monitors the solar activity generating binary files from which solar maps can be extracted. The identification of Active Regions (AR) in these maps is affected by the strong atmospheric attenuation and inaccuracies of the telescope pointing, therefore, maps are visually inspected to manually extract the AR. This is a lengthy process if one wants to do statistical analysis over the 20 years data set already recorded. To automate the process, we propose artificial intelligence techniques of machine learning and computer vision. A Convolutional Neural Network was created within the Keras framework for the classification of the SST maps and then, a computer vision algorithm in the OpenCV framework for the automatic detection of AR. This hybrid approach allowed the identification of more than 400 active regions between January 2002 and December 2017 and the statistical analysis of their physical properties. Our results were validated by comparing with previous works which were carried out with a visual identification and manual extraction procedure, and we found good agreement.
SST, Submillimeter Solar Maps, Artificial Intelligence, Convolutional Neural Networks, Keras, Computational View, OpenCV.