Weihao Shen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Weihao Shen has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Yanfei Zhong
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
4 shared publications - Ailong Ma
Wuhan University
3 shared publications - Junjue Wang
The University of Tokyo
3 shared publications - Zhuo Zheng
Department of Computer Science, Stanford University, Stanford, CA, USA
2 shared publications - Ruiyi Yang
Wuhan University
1 shared publication - Tian Ke
State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan, China
1 shared publication
Latest Publications
- 10 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 4705314
Martian Mineral Formation Analysis Based on Martian Mineral Corpus and Knowledge Graph
Martian minerals are key indicators of the planet’s geological evolution and potential habitability. Their formation mechanisms record critical evidence of early aqueous activity, surface geochemistry, and geological activities. However, analyzing these mechanisms requires integrating heterogeneous data from spectroscopy, geomorphology, and stratigraphy. Existing studies still rely heavily on manual interpretation of heterogeneous data, resulting in inconsistent conclusions and a time-consuming reasoning process. To address this challenge, we build the first unified Martian Mineral Corpus and Martian Mineral Knowledge Graph (MMKG). The corpus integrates global Mars data products and scientific literature, while the MMKG unifies mineral-related concepts with their semantic and spatiotemporal relations, providing comprehensive data and knowledge support for reasoning. On this basis, RBG2 is proposed as an intelligent question answering (QA) framework designed for Martian mineral analysis. RBG2 performs knowledge graph path retrieval under dual constraints: geological vectors distilled from the unified corpus guide multi-hop exploration, and geological knowledge-consistency validates candidate chains. This yields graph paths that are both context-aligned and causally plausible. Leveraging these paths as the backbone, RBG2 integrates geological context from the corpus and the knowledge graph to perform unified reasoning. Experimental results show that RBG2 outperforms existing methods on Martian mineral question answering. For mineral formation mechanisms, RBG2 covers over 90% of key points of experts’ analysis on average and produces reasoning that is both interpretable and traceable, achieving the first reliable and intelligent Martian mineral analysis. Code has been made available at: https://github.com/guoluqingfeng10086/MMQA.
- 10 June 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 64 (01962892) , 4705314
Martian Mineral Formation Analysis Based on Martian Mineral Corpus and Knowledge Graph
Martian minerals are key indicators of the planet’s geological evolution and potential habitability. Their formation mechanisms record critical evidence of early aqueous activity, surface geochemistry, and geological activities. However, analyzing these mechanisms requires integrating heterogeneous data from spectroscopy, geomorphology, and stratigraphy. Existing studies still rely heavily on manual interpretation of heterogeneous data, resulting in inconsistent conclusions and a time-consuming reasoning process. To address this challenge, we build the first unified Martian Mineral Corpus and Martian Mineral Knowledge Graph (MMKG). The corpus integrates global Mars data products and scientific literature, while the MMKG unifies mineral-related concepts with their semantic and spatiotemporal relations, providing comprehensive data and knowledge support for reasoning. On this basis, RBG2 is proposed as an intelligent question answering (QA) framework designed for Martian mineral analysis. RBG2 performs knowledge graph path retrieval under dual constraints: geological vectors distilled from the unified corpus guide multi-hop exploration, and geological knowledge-consistency validates candidate chains. This yields graph paths that are both context-aligned and causally plausible. Leveraging these paths as the backbone, RBG2 integrates geological context from the corpus and the knowledge graph to perform unified reasoning. Experimental results show that RBG2 outperforms existing methods on Martian mineral question answering. For mineral formation mechanisms, RBG2 covers over 90% of key points of experts’ analysis on average and produces reasoning that is both interpretable and traceable, achieving the first reliable and intelligent Martian mineral analysis. Code has been made available at: https://github.com/guoluqingfeng10086/MMQA.
- Open Access
- 17 March 2026
- Published by Taylor & Francis in Geo-spatial Information Science
- p. 1-19
Visual state space models with spiral selective scan for referring remote-sensing image segmentation
- 17 March 2026
- Published by Taylor & Francis in Geo-spatial Information Science
- p. 1-19
Open AccessVisual state space models with spiral selective scan for referring remote-sensing image segmentation
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
Adaptive Self-Supporting Prototype Learning for Remote Sensing Few-Shot Semantic Segmentation
The semantic segmentation of remote sensing images with few shots has important theoretical and application value. Most of the existing few-shot semantic segmentation frameworks are based on prototype learning methods, in which a single support prototype is designed to guide the query set for prediction. However, the visual differences between the support set and the query set make it difficult for a single support prototype, generated from the support set, to comprehensively encapsulate the semantic information of all the query images. This paper introduces an adaptive self-supporting prototype learning network designed for few-shot segmentation (FSS), in order to tackle the challenges mentioned earlier. We propose adaptive hyper-prototype representation (HPR), which consists of hyper-prototype clustering (HPC) and guided prototype matching (GPM), to generate and assign multiple representative prototypes to compensate for the limitations of a single prototype in representing the semantic information of the query images. Specifically, HPC is a parameter-free and adaptive approach, which can extract more representative prototypes by aggregating similar feature vectors utilizing superpixel feature clustering. Meanwhile, GPM can select matched prototypes to provide more accurate guidance, allowing for uniformly aligned representation of multiple prototypes and complex image semantic information. We also introduce self-supporting matching (SSM) prototype learning, which can accurately guide the query set segmentation by acquiring query set prototypes. SSM generates initial pseudo-labels for the query set based on the support set prototypes, and further guides the query set using the pseudo-labels, along with the query prototypes generated by its own features, thus effectively avoiding visual differences between the support set and query set. The proposed adaptive self-supporting prototype learning network substantially improves the prototype quality and achieves a superior performance on object-level remote sensing datasets.
- 29 July 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Geoscience and Remote Sensing
- Vol. 62 (01962892) , 1-16
Adaptive Self-Supporting Prototype Learning for Remote Sensing Few-Shot Semantic Segmentation
The semantic segmentation of remote sensing images with few shots has important theoretical and application value. Most of the existing few-shot semantic segmentation frameworks are based on prototype learning methods, in which a single support prototype is designed to guide the query set for prediction. However, the visual differences between the support set and the query set make it difficult for a single support prototype, generated from the support set, to comprehensively encapsulate the semantic information of all the query images. This paper introduces an adaptive self-supporting prototype learning network designed for few-shot segmentation (FSS), in order to tackle the challenges mentioned earlier. We propose adaptive hyper-prototype representation (HPR), which consists of hyper-prototype clustering (HPC) and guided prototype matching (GPM), to generate and assign multiple representative prototypes to compensate for the limitations of a single prototype in representing the semantic information of the query images. Specifically, HPC is a parameter-free and adaptive approach, which can extract more representative prototypes by aggregating similar feature vectors utilizing superpixel feature clustering. Meanwhile, GPM can select matched prototypes to provide more accurate guidance, allowing for uniformly aligned representation of multiple prototypes and complex image semantic information. We also introduce self-supporting matching (SSM) prototype learning, which can accurately guide the query set segmentation by acquiring query set prototypes. SSM generates initial pseudo-labels for the query set based on the support set prototypes, and further guides the query set using the pseudo-labels, along with the query prototypes generated by its own features, thus effectively avoiding visual differences between the support set and query set. The proposed adaptive self-supporting prototype learning network substantially improves the prototype quality and achieves a superior performance on object-level remote sensing datasets.