Abstract: Aiming at the problems of difficult observation, high acquisition cost and different time scales for the input data of traditional auroral electrojet (AE) index prediction models, this paper proposes a prediction model based on ultraviolet aurora images. In this paper, the ultraviolet aurora images of the Polar satellite are used as the data base, and the longitude and latitude distribution characteristics of Aurora intensity is extracted by the grid feature extraction method, which are input into the generalized regression neural network to predict the AE index. Based on the research on the relationship between the auroral electrojet system and the spatial distribution of auroral power, the contribution of the geomagnetic latitude and longitude distribution characteristics of auroral power to the AE index was further explored. The experimental results of the prediction model show that the method of predicting the AE index with the aurora image data is feasible, and compared with the traditional prediction model, the model in this paper is superior to the traditional prediction model in the evaluation criteria of root mean square error RMSE, average relative variance ARV and determination coefficient R².