Advances in Transformer-Based Models for Multi-Modal Remote Sensing Data Fusion and Analysis

Dear Colleagues,

Rapid advancements in Earth observation technology are driving cutting-edge research in the field of remote sensing. Multi-source, multi-modal remote sensing data—acquired from diverse sensors, such as optical, Synthetic Aperture Radar, Light Detection and Ranging, and hyperspectral sensors—offer unprecedented insights and observational perspectives in Earth system science. Consequently, the effective fusion of these heterogeneous data, which possess distinct physical characteristics and data structures, and the synergistic extraction of high-level, high-precision semantic information represent key scientific challenges in remote sensing information processing.

As a landmark achievement in deep learning, the Transformer model, through its core self-attention mechanism, exhibits an exceptional ability to capture long-range dependencies and global contextual information. This provides a novel paradigm for the deep fusion and synergistic interpretation of multi-modal remote sensing data. This paradigm has the potential to overcome the performance limitations of traditional methods, thereby significantly enhancing the accuracy and sophistication of typical remote sensing applications, including land cover classification, object detection, change detection, and scene understanding. Therefore, this systematic investigation and review of Transformer-based methods for the fusion and interpretation of multi-modal remote sensing data holds both significant theoretical importance and broad practical value.

This Special Issue aims to highlight cutting-edge theories, key technologies, and innovative applications of Transformer models in multi-modal remote sensing data fusion and analysis. We invite experts and scholars in both academia and industry to submit original research, comprehensive reviews, and perspectives on future directions in this field.

The scope of this Special Issue includes, but is not limited to, the following topics:

1. Innovations in Transformer Model Architectures

• Cross-modal attention models for the deep fusion of multi-modal remote sensing data.
• Lightweight and efficient Transformer model designs for large-scale remote sensing data processing.
• Hybrid model architectures that fuse Transformers with other deep learning networks.

2. Applications of Transformer-based Intelligent Interpretation for Multi-modal Remote Sensing

• High-precision Land Use/Land Cover mapping.
• Object detection, recognition, and semantic segmentation in complex scenes.
• Change detection methods fusing multi-temporal and multi-modal data.
• Classification, segmentation, and scene understanding of 3D point-cloud data.
• Remote sensing image quality enhancement, such as image inpainting and super-resolution reconstruction.
• Demonstration and validation of applications in typical domains such as agriculture, forestry, hydrology, and disaster emergency response.

3. Model Theory, Training, and Optimization

• Development and fine-tuning techniques for large-scale pre-trained models tailored to the remote sensing domain.
• Self-supervised, unsupervised, or few-shot learning paradigms and methods suitable for the unique characteristics of remote sensing data.
• Studies of the interpretability of Transformer models in remote sensing applications.

Dr. Bowen Du
Dr. Hongming Zhu
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Remote Sensing is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2700 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• remote sensing foundation models
• vision transformer
• multi-scale transformer
• efficient transformer
• cross-modal attention
• multi-scale feature fusion
• spatiotemporal data fusion
• heterogeneous data fusion
• multi-scale transformer
• multi-modal representation
• cross-modal representation learning
• multi-modal semantic segmentation
• multimodal scene understanding