“Multimodal classification tasks have made new breakthroughs in the field of remote sensing image processing. In response to the challenges of fusing hyperspectral images (HSI) and LiDAR data, researchers have proposed a CNN Transformer collaborative classification network based on contrastive learning called CLCT Net. The network realizes semantic alignment between different sensor data through an innovative common feature extraction module, and solves the problems of cross modal information representation and feature alignment. The researchers designed a dual branch HSI encoder that includes spatial channel branches and spectral context branches, as well as a LiDAR encoder that combines frequency domain self attention mechanism, in order to fully extract and integrate the rich features of the two types of data. Meanwhile, they ingeniously integrated contrastive learning methods to further improve the accuracy of multimodal data collaborative classification. Experimental verification on the Houston 2013 and Trento datasets showed that the land cover classification accuracy of the CLCT Net model reached 92.01% and 98.90%, respectively, significantly surpassing other similar models. This achievement not only provides powerful tools for deep mining and collaborative extraction of cross modal data features, but also opens up new directions for research on multimodal classification tasks in the field of remote sensing image processing.”