Local Information Interaction Transformer for Hyperspectral and LiDAR Data Classification

The multisource remote sensing classification task has two main challenges. 1) How to capture hyperspectral image (HSI) and light detection ranging (LiDAR) features cooperatively fully mine the complementary information between data. 2) adaptively fuse features, which should not only overcome imbalance HSI LiDAR data but also avoid generation of redundant information. local interaction transformer (LIIT) model proposed herein can effectively address these above issues. Specifically, multibranch feature embedding is first performed help in fine-grained serialization features; subsequently, a local-based interactor (L-MSFI) designed explore together. This structure provides an transmission environment for further alleviates homogenization processing mode self-attention process. More importantly, selection module (MSTSM) developed dynamically solve problem insufficient fusion. Experiments were carried out on three remote-sensing datasets, results show that LIIT more performance advantages than state-of-the-art CNN methods.