A Novel Transformer-KAN Network Utilizes Multimodality Contrastive Learning and Mask Reconstruction for Remote Sensing Classification

Contrastive learning is a promising paradigm in the multimodal few-shot domain, leveraging images captured from different sensors as diverse perspectives of the same ground objects. However, existing methods encounter two major challenges: relatively weak discriminative ability, complex features inherent in remote sensing data, and insufficient cross-modal interactions, as these methods often align features within individual modalities without fully exploiting their complementary information. These limitations hinder the overall performance of models in multimodal remote sensing tasks. In this article, we propose a novel framework to address these challenges. First, we introduce the Transformer as the encoder, and at the same time, we replace traditional multilayer perceptrons (MLPs) with the Kolmogorov–Arnold network (KAN) for feature dimensionality reduction. The model reduces parameters slightly without affecting accuracy, achieving better efficiency. Second, we design a specialized projection layer to align and deeply integrate spatial and light detection and ranging (LiDAR) modalities, effectively harnessing their complementary strengths and implementing a self-supervised decoding mechanism to reconstruct masked image patches generated during the data augmentation process, enforcing spatial and structural learning for robust multimodal representations. Extensive experiments conducted on four public datasets demonstrate the superior classification performance of our proposed method, outperforming state-of-the-art approaches and confirming its efficacy in multimodal remote sensing tasks.