-MAE: A spatial-temporal-spectral unified remote sensing pre-training method based on anchor-aware masked autoencoder.