TVLightFormer: A Lightweight Cross-Modal Transformer for Language-Guided Target Localization in SAR Imagery
Remote Sensing · 04.05.2026 · 18 (9) · 1430
DOI: 10.3390/rs18091430 ↗Kurz gefasst
Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 04.05.2026. Verfasst von Yuqiao Zhong und 4 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Undergraduate Training Program for Innovation and Entrepreneurship, National Natural Science Foundation of China, Excellent Young Scientist Fund of Hunan Province.
Abstract (Originalsprache)
We study language-guided target localization in synthetic aperture radar (SAR) imagery for deployment on resource-constrained platforms. Existing vision-language models either rely on heavy backbones unsuitable for edge devices or are designed for natural images, overlooking SAR-specific characteristics such as speckle noise, weak scattering responses, and geometric distortions. The proposed model, TVLightFormer, combines a lightweight dual-modal encoder (MobileNetV3 and TinyBERT) with a grouped-query attention (GQA) mechanism for efficient cross-modal interaction and an activation-free lightweight feature pyramid network (LFPN) to handle scale variation while preserving weak scattering signals. The individual modules are not claimed as newly invented components; the main contribution lies in their SAR-aware integration for edge-oriented cross-modal localization. We evaluate the model on five remote sensing datasets—SOMA-1M, ATRNet-STAR, GAIA, MLRSNet, and SODAS—under a unified localization setting, and we explicitly discuss the limitations introduced by weak or scene-level annotations. The results show that TVLightFormer achieves a favorable trade-off between accuracy and efficiency, reaching an average mIoU of 69.8% with 27.4 M parameters and 9.7 GFLOPs. Ablation studies quantify the contribution of each component. The model is suited for edge-oriented scenarios where computational resources are limited. We also provide a critical analysis of failure cases, SAR-specific disturbance factors, loss-function choices, and dataset-protocol sensitivity.
Förderung
- National Undergraduate Training Program for Innovation and Entrepreneurship
- National Natural Science Foundation of China
- Excellent Young Scientist Fund of Hunan Province
Zitieren
APA
Zhong, Y., Quan, H., Nie, C., Wei, Y., & Guo, Y. (2026). TVLightFormer: A Lightweight Cross-Modal Transformer for Language-Guided Target Localization in SAR Imagery. Remote Sensing, 18(9), 1430. https://doi.org/10.3390/rs18091430
Vancouver
Zhong Y, Quan H, Nie C, Wei Y, Guo Y. TVLightFormer: A Lightweight Cross-Modal Transformer for Language-Guided Target Localization in SAR Imagery. Remote Sensing. 2026;18(9):1430. doi:10.3390/rs18091430
BibTeX
@article{Zhong2026,
author = {Zhong, Yuqiao and Quan, Haoqi and Nie, Chenyu and Wei, Yingmei and Guo, Yanming},
title = {TVLightFormer: A Lightweight Cross-Modal Transformer for Language-Guided Target Localization in SAR Imagery},
journal = {Remote Sensing},
year = {2026},
volume = {18},
number = {9},
pages = {1430},
doi = {10.3390/rs18091430}
}
EUVIMED