Our framework for enhancing unsupervised classification and schematic diagram for feature extraction. Panel a represents the improved model flowchart, while Panel b represents the schematic of feature extraction. In the feature extraction process, we use the ConNeXt large model for data encoding to extract compelling features from the data. Subsequently, the next step involves PCA dimensionality reduction to eliminate redundant information in the image. This combination effectively extracts features while retaining only basic information, optimizing computing resources.