Geomagnetic storms are the most important type of geomagnetic field disturbance, affecting fields such as communication, power, and aerospace. Based on the minute value data of fixed-point geomagnetic observations, spectral imaging technology is used, and the VGG19 convolutional neural network model is used to carry out artificial intelligence image classification research on "geomagnetic storm days" and "geomagnetic quiet days". The results indicate that spectrograms have high practicality for image recognition and classification, and the VGG19 convolutional neural network model is highly feasible for classifying geomagnetic "geomagnetic storm days" and "geomagnetic quiet days".