Enhancing hyperspectral image classification: Leveraging unsupervised information with guided group contrastive learning