2 July 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
With the rapid advancement of deep learning technologies, the use of computers for processing and analyzing multimodal remote sensing data has attracted significant attention. Joint land cover classification (LCC) using hyperspectral image (HSI) and light detection and ranging (LiDAR) data is one such key area of research. Recently, the emerging deep learning framework Mamba has shown superior performance over traditional architectures, including transformers and convolutional neural networks. However, its application to LCC faces challenges. First, existing methods typically use a unified feature extraction approach for heterogeneous data, which fails to capture the distinct characteristics of each data modality. Moreover, effectively integrating these heterogeneous features remains a significant difficulty. To address these issues, we propose the multimodal fusion Mamba network (M2FMNet), which consists of three key components: the spatial–spectral adaptive Mamba, the elevation-enhanced Mamba, and an enhanced fusion module. The spatial–spectral adaptive Mamba combines two distinct scanning methods to efficiently extract spatial and spectral features from HSI data. The elevation-enhanced Mamba, designed for LiDAR data, utilizes a specialized scanning method to extract elevation features. Finally, the enhanced fusion module integrates intra- and intermodal relationships of these features at the pixel level. Experimental results on three real-world datasets show that the M2FMNet outperforms state-of-the-art methods in both qualitative and quantitative metrics.
1 May 2025 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Joint hyperspectral (HS) and Light Detection and Ranging (LiDAR) data for land cover classification have received widespread attention. Nevertheless, existing methods often struggle with multi-scale feature extraction and multi-source feature fusion. To address these issues, this paper proposes a pyramid progressive cross-fusion network, named PyionNet, for joint classification of HS and LiDAR data. Technically, the proposed PyionNet incorporates a pyramid multi-scale extraction (PME) module cascaded with a progressive cross-fusion (PCF) module, which are designed to extract multi-scale features and fuse multi-source features, respectively. Specifically, the PME module employs pyramid spectral and spatial convolutional blocks to extract and aggregate features from HS and LiDAR data, respectively, then uses an expansion-compression strategy to eliminate redundant information, and thus achieve a compact multi-scale feature representation. Moreover, the PCF module utilizes the transformer architecture and introduces two innovative feature fusion blocks, i.e., the discrepancy feature fusion (DFF) block and the common feature fusion (CFF) block, to progressively integrate multi-source and multi-scale features extracted by the PME module. The DFF block is designed to integrate discrepancy features between HS and LiDAR data to enhance complementary information. Subsequently, the CFF block focuses on fusing common features to progressively enrich the fused representation of the DFF block and further enhance the representation power. Experiments on three real-world datasets demonstrate that the proposed PyionNet outperforms state-of-the-art methods in both qualitative and quantitative results.