RHG-DETR: Riemannian Hyper-Graph Transformer with Dynamic Receptive Fields for Detecting Special Targets in Degraded UAV Imagery.
Saved in:
| Title: | RHG-DETR: Riemannian Hyper-Graph Transformer with Dynamic Receptive Fields for Detecting Special Targets in Degraded UAV Imagery. |
|---|---|
| Authors: | Wang, Kaipeng1 (AUTHOR), He, Guanglin1 (AUTHOR) heguanglin@bit.edu.cn, Kong, Wenhao1 (AUTHOR), Fu, Yuzhe1 (AUTHOR), Li, Zongze1 (AUTHOR) |
| Source: | Remote Sensing. Jun2026, Vol. 18 Issue 11, p1728. 29p. |
| Subjects: | Automatic target recognition, Feature extraction, Drone photography, Transformer models, Object recognition (Computer vision) |
| Abstract: | Highlights: What are the main findings? RHG-DETR is a novel detection framework for special target detection under multi-type degradation in UAV remote sensing imagery. Three synergistic innovation modules (DRHANet, BWAFN, ASMED) respectively enhance multi-scale feature extraction, cross-scale semantic fusion, and sparse encoding capability, while effectively suppressing background interference. On a self-constructed special target dataset, RHG-DETR achieves an mAP50 of 78.5%, surpassing the baseline model RT-DETR and other mainstream methods, while reducing computational cost and parameter count by 34.4% and 28.8%, respectively, at an inference speed of 84.2 FPS. What are the implications of the main findings? Effective modeling of robustness to multi-type degradation (blur, rain, snow, fog, low illumination, strong light, and electromagnetic interference) is validated as a key factor in significantly improving UAV special target detection performance, breaking through the limitations of existing detectors under standard clear imaging conditions. This framework provides a lightweight and practical solution for real-time special target perception under resource-constrained conditions, with broad application potential in all-weather autonomous driving, disaster rescue, and infrastructure inspection. Special target detection in UAV remote sensing imagery is challenged by composite multi-type degradation, which collectively erodes target structure across every stage of a detection pipeline. Existing methods address individual degradation types in isolation and do not generalize to the composite conditions encountered in real deployment. We propose the Riemannian Hyper-Graph Detection Transformer (RHG-DETR), a degradation-robust end-to-end framework composed of the Dynamic Receptive-field Hyper-graph Attention Network (DRHANet), the Bi-directional Weighted Adaptive Fusion Network (BWAFN), and the Adaptive Sparse Multi-scale Encoder with Dynamic Normalization (ASMED). DRHANet introduces anisotropic dynamic depthwise separable convolutions to align receptive fields with local structural orientations and Riemannian hyper-graph fusion to aggregate multi-scale features on a manifold, preserving inter-scale angular relations that Euclidean fusion destroys under degradation. BWAFN employs a bi-directional weighted pyramid in which each fusion node learns per-scale contribution weights, correcting cross-scale semantic misalignment that fixed-weight single-pass aggregation cannot recover. ASMED combines saliency-conditioned sparse window attention to suppress background dilution, a spatially gated feed-forward branch to retain pre-attention spatial geometry, and a bounded dynamic normalizer to stabilize activations under extreme illumination and electromagnetic interference. On a self-constructed UAV special-target dataset spanning seven physics-based degradation types, RHG-DETR achieves 78.5% mAP50, a 3.7% absolute gain over RT-DETR at 34.4% lower GFLOPs and 28.8% fewer parameters at 84.2 FPS, outperforming restoration-then-detect pipelines in both accuracy and latency. Consistent improvements on VisDrone2019 and BDD100K confirm cross-domain generalization. [ABSTRACT FROM AUTHOR] |
| Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| FullText | Links: – Type: pdflink Text: Availability: 1 |
|---|---|
| Header | DbId: egs DbLabel: Engineering Source An: 194586949 AccessLevel: 6 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 0 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: RHG-DETR: Riemannian Hyper-Graph Transformer with Dynamic Receptive Fields for Detecting Special Targets in Degraded UAV Imagery. – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Wang%2C+Kaipeng%22">Wang, Kaipeng</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22He%2C+Guanglin%22">He, Guanglin</searchLink><relatesTo>1</relatesTo> (AUTHOR)<i> heguanglin@bit.edu.cn</i><br /><searchLink fieldCode="AR" term="%22Kong%2C+Wenhao%22">Kong, Wenhao</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Fu%2C+Yuzhe%22">Fu, Yuzhe</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Li%2C+Zongze%22">Li, Zongze</searchLink><relatesTo>1</relatesTo> (AUTHOR) – Name: TitleSource Label: Source Group: Src Data: <searchLink fieldCode="JN" term="%22Remote+Sensing%22">Remote Sensing</searchLink>. Jun2026, Vol. 18 Issue 11, p1728. 29p. – Name: Subject Label: Subjects Group: Su Data: <searchLink fieldCode="DE" term="%22Automatic+target+recognition%22">Automatic target recognition</searchLink><br /><searchLink fieldCode="DE" term="%22Feature+extraction%22">Feature extraction</searchLink><br /><searchLink fieldCode="DE" term="%22Drone+photography%22">Drone photography</searchLink><br /><searchLink fieldCode="DE" term="%22Transformer+models%22">Transformer models</searchLink><br /><searchLink fieldCode="DE" term="%22Object+recognition+%28Computer+vision%29%22">Object recognition (Computer vision)</searchLink> – Name: Abstract Label: Abstract Group: Ab Data: Highlights: What are the main findings? RHG-DETR is a novel detection framework for special target detection under multi-type degradation in UAV remote sensing imagery. Three synergistic innovation modules (DRHANet, BWAFN, ASMED) respectively enhance multi-scale feature extraction, cross-scale semantic fusion, and sparse encoding capability, while effectively suppressing background interference. On a self-constructed special target dataset, RHG-DETR achieves an mAP50 of 78.5%, surpassing the baseline model RT-DETR and other mainstream methods, while reducing computational cost and parameter count by 34.4% and 28.8%, respectively, at an inference speed of 84.2 FPS. What are the implications of the main findings? Effective modeling of robustness to multi-type degradation (blur, rain, snow, fog, low illumination, strong light, and electromagnetic interference) is validated as a key factor in significantly improving UAV special target detection performance, breaking through the limitations of existing detectors under standard clear imaging conditions. This framework provides a lightweight and practical solution for real-time special target perception under resource-constrained conditions, with broad application potential in all-weather autonomous driving, disaster rescue, and infrastructure inspection. Special target detection in UAV remote sensing imagery is challenged by composite multi-type degradation, which collectively erodes target structure across every stage of a detection pipeline. Existing methods address individual degradation types in isolation and do not generalize to the composite conditions encountered in real deployment. We propose the Riemannian Hyper-Graph Detection Transformer (RHG-DETR), a degradation-robust end-to-end framework composed of the Dynamic Receptive-field Hyper-graph Attention Network (DRHANet), the Bi-directional Weighted Adaptive Fusion Network (BWAFN), and the Adaptive Sparse Multi-scale Encoder with Dynamic Normalization (ASMED). DRHANet introduces anisotropic dynamic depthwise separable convolutions to align receptive fields with local structural orientations and Riemannian hyper-graph fusion to aggregate multi-scale features on a manifold, preserving inter-scale angular relations that Euclidean fusion destroys under degradation. BWAFN employs a bi-directional weighted pyramid in which each fusion node learns per-scale contribution weights, correcting cross-scale semantic misalignment that fixed-weight single-pass aggregation cannot recover. ASMED combines saliency-conditioned sparse window attention to suppress background dilution, a spatially gated feed-forward branch to retain pre-attention spatial geometry, and a bounded dynamic normalizer to stabilize activations under extreme illumination and electromagnetic interference. On a self-constructed UAV special-target dataset spanning seven physics-based degradation types, RHG-DETR achieves 78.5% mAP50, a 3.7% absolute gain over RT-DETR at 34.4% lower GFLOPs and 28.8% fewer parameters at 84.2 FPS, outperforming restoration-then-detect pipelines in both accuracy and latency. Consistent improvements on VisDrone2019 and BDD100K confirm cross-domain generalization. [ABSTRACT FROM AUTHOR] – Name: AbstractSuppliedCopyright Label: Group: Ab Data: <i>Copyright of Remote Sensing is the property of MDPI and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.) |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=194586949 |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.3390/rs18111728 Languages: – Code: eng Text: English PhysicalDescription: Pagination: PageCount: 29 StartPage: 1728 Subjects: – SubjectFull: Automatic target recognition Type: general – SubjectFull: Feature extraction Type: general – SubjectFull: Drone photography Type: general – SubjectFull: Transformer models Type: general – SubjectFull: Object recognition (Computer vision) Type: general Titles: – TitleFull: RHG-DETR: Riemannian Hyper-Graph Transformer with Dynamic Receptive Fields for Detecting Special Targets in Degraded UAV Imagery. Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Wang, Kaipeng – PersonEntity: Name: NameFull: He, Guanglin – PersonEntity: Name: NameFull: Kong, Wenhao – PersonEntity: Name: NameFull: Fu, Yuzhe – PersonEntity: Name: NameFull: Li, Zongze IsPartOfRelationships: – BibEntity: Dates: – D: 01 M: 06 Text: Jun2026 Type: published Y: 2026 Identifiers: – Type: issn-print Value: 20724292 Numbering: – Type: volume Value: 18 – Type: issue Value: 11 Titles: – TitleFull: Remote Sensing Type: main |
| ResultId | 1 |
