We present metadata (date and time of observations) extracted from the digitized (scanned) photographic images of solar disk observations in neutral hydrogen Hα spectral line made at the U.S. National Solar Observatory at Sacramento Peak during 1966–2002. Successful extraction of metadata enables a more effective use of this historical data set in scientific research. About 4.8 million of images were processed using machine-learning models trained to detect the text fields and digitize text information in these fields. We report digitized data along with confidence estimates than can be used to filter out less reliable data. We discuss the complications in digitizing this historical dataset, and suggest additional steps that could be used to improve the extraction of metadata.
Illarionov et al. (Mon,) studied this question.