Abstract
Multimodal remote sensing data, such as hyperspectral and LiDAR imagery, provide complementary information for land cover analysis. However, effectively clustering these heterogeneous yet spatially aligned data remains challenging due to cross-modal inconsistency and data complexity. In this work, we propose an anchor-guided multi-view fuzzy clustering (AMVFC) framework to achieve robust and consistent clustering across multiple modalities. The proposed approach represents cluster structures through a set of anchor points and incorporates a shared fuzzy membership to promote cross-modal consistency, while preserving the characteristics of each modality. Furthermore, a deep extension of the framework is developed to better capture nonlinear relationships in multimodal data. Experiments on three benchmark datasets demonstrate that the proposed methods achieve competitive and consistently improved clustering performance compared with existing approaches. Our code is available at https://github.com/kcarol1/AMVFC.