HybriDet: A Hybrid Neural Network Combining CNN and Transformer for Wildfire Detection in Remote Sensing Imagery
I tiakina i:
| I whakaputaina i: | Remote Sensing vol. 17, no. 20 (2025), p. 3497-3520 |
|---|---|
| Kaituhi matua: | |
| Ētahi atu kaituhi: | |
| I whakaputaina: |
MDPI AG
|
| Ngā marau: | |
| Urunga tuihono: | Citation/Abstract Full Text + Graphics Full Text - PDF |
| Ngā Tūtohu: |
Kāore He Tūtohu, Me noho koe te mea tuatahi ki te tūtohu i tēnei pūkete!
|
MARC
| LEADER | 00000nab a2200000uu 4500 | ||
|---|---|---|---|
| 001 | 3265942939 | ||
| 003 | UK-CbPIL | ||
| 022 | |a 2072-4292 | ||
| 024 | 7 | |a 10.3390/rs17203497 |2 doi | |
| 035 | |a 3265942939 | ||
| 045 | 2 | |b d20250101 |b d20251231 | |
| 084 | |a 231556 |2 nlm | ||
| 100 | 1 | |a Dong Fengming |u School of Mechanical, Electrical & Information Engineering, Shandong University, Weihai 264209, China; 202100800012@mail.sdu.edu.cn | |
| 245 | 1 | |a HybriDet: A Hybrid Neural Network Combining CNN and Transformer for Wildfire Detection in Remote Sensing Imagery | |
| 260 | |b MDPI AG |c 2025 | ||
| 513 | |a Journal Article | ||
| 520 | 3 | |a <sec sec-type="highlights"> What are the main findings? <list list-type="bullet"> <list-item> </list-item>A novel hybrid neural network architecture named HybriDet is proposed, which effectively integrates the local feature extraction capability of CNNs and the global contextual modeling strength of Transformers. The innovative SwinBottle module and Coordinate-Spatial (CS) dual attention mechanism significantly improve the detection accuracy for wildfires and smoke in complex remote sensing imagery. <list-item> A superior balance between accuracy and efficiency is achieved. The lightweight model after structured pruning contains only 6.45 M parameters. It significantly outperforms state-of-the-art models like YOLOv8 by 6.4% in mAP50 on the FASDD-RS dataset while maintaining real-time inference speed suitable for edge device deployment. </list-item> What are the implications of the main findings? <list list-type="bullet"> <list-item> </list-item>Provides an efficient and reliable fire detection solution for resource-constrained edge computing environments (e.g., satellites, UAVs). Model compression and optimization techniques enable the practical deployment of high-performance deep learning models on low-power devices, directly contributing to early wildfire warning and emergency response. <list-item> The proposed method demonstrates strong generalization capabilities and broad application prospects. Its superior performance across multiple public datasets (FASDD-UAV, FASDD-RS, VOC) indicates its effectiveness in handling highly heterogeneous remote sensing imagery, providing crucial technical support for intelligent remote sensing monitoring in ecological conservation and socioeconomic security. </list-item> Early warning systems on edge devices such as satellites and unmanned aerial vehicles (UAVs) are essential for effective forest fire prevention. Edge Intelligence (EI) enables deploying deep learning models on edge devices; however, traditional convolutional neural networks (CNNs)/Transformer-based models struggle to balance local-global context integration and computational efficiency in such constrained environments. To address these challenges, this paper proposes HybriDet, a novel hybrid-architecture neural network for wildfire detection. This architecture integrates the strengths of both CNNs and Transformers to effectively capture both local and global contextual information. Furthermore, we introduce efficient attention mechanisms—Windowed Attention and Coordinate-Spatial (CS) Attention—to simultaneously enhance channel-wise and spatial-wise features in high-resolution imagery, enabling long-range dependency modeling and discriminative feature extraction. Additionally, to optimize deployment efficiency, we also apply model pruning techniques to improve generalization performance and inference speed. Extensive experimental evaluations demonstrate that HybriDet achieves superior feature extraction capabilities while maintaining high computational efficiency. The optimized lightweight variant of HybriDet has a compact model size of merely 6.45 M parameters, facilitating seamless deployment on resource-constrained edge devices. Comparative evaluations on the FASDD-UAV, FASDD-RS, and VOC datasets demonstrate that HybriDet achieves superior performance over state-of-the-art models, particularly in processing highly heterogeneous remote sensing (RS) imagery. When benchmarked against YOLOv8, HybriDet demonstrates a 6.4% enhancement in mAP50 on the FASDD-RS dataset while maintaining comparable computational complexity. Meanwhile, on the VOC dataset and the FASDD-UAV dataset, our model improved by 3.6% and 0.2%, respectively, compared to the baseline model YOLOv8. These advancements highlight HybriDet’s theoretical significance as a novel hybrid EI framework for wildfire detection, with practical implications for disaster emergency response, socioeconomic security, and ecological conservation. | |
| 653 | |a Feature extraction | ||
| 653 | |a Forest fires | ||
| 653 | |a Warning systems | ||
| 653 | |a Datasets | ||
| 653 | |a Computer architecture | ||
| 653 | |a Forest fire detection | ||
| 653 | |a Early warning systems | ||
| 653 | |a Artificial neural networks | ||
| 653 | |a Edge computing | ||
| 653 | |a Unmanned aerial vehicles | ||
| 653 | |a Computer applications | ||
| 653 | |a Machine learning | ||
| 653 | |a Emergency preparedness | ||
| 653 | |a Wildfires | ||
| 653 | |a Remote sensing | ||
| 653 | |a Computer vision | ||
| 653 | |a Optimization | ||
| 653 | |a Effectiveness | ||
| 653 | |a Computational efficiency | ||
| 653 | |a Socioeconomics | ||
| 653 | |a Algorithms | ||
| 653 | |a Conservation | ||
| 653 | |a Accuracy | ||
| 653 | |a Deep learning | ||
| 653 | |a Image resolution | ||
| 653 | |a Models | ||
| 653 | |a Disaster management | ||
| 653 | |a Remote monitoring | ||
| 653 | |a Satellites | ||
| 653 | |a Fire prevention | ||
| 653 | |a Embedded systems | ||
| 653 | |a Neural networks | ||
| 653 | |a Design | ||
| 653 | |a Devices | ||
| 653 | |a Complexity | ||
| 653 | |a Constraints | ||
| 653 | |a Parameters | ||
| 653 | |a Cybersecurity | ||
| 700 | 1 | |a Wang, Ming |u Inspur Cloud Information Technology Co., Ltd., Jinan 250101, China | |
| 773 | 0 | |t Remote Sensing |g vol. 17, no. 20 (2025), p. 3497-3520 | |
| 786 | 0 | |d ProQuest |t Advanced Technologies & Aerospace Database | |
| 856 | 4 | 1 | |3 Citation/Abstract |u https://www.proquest.com/docview/3265942939/abstract/embedded/L8HZQI7Z43R0LA5T?source=fedsrch |
| 856 | 4 | 0 | |3 Full Text + Graphics |u https://www.proquest.com/docview/3265942939/fulltextwithgraphics/embedded/L8HZQI7Z43R0LA5T?source=fedsrch |
| 856 | 4 | 0 | |3 Full Text - PDF |u https://www.proquest.com/docview/3265942939/fulltextPDF/embedded/L8HZQI7Z43R0LA5T?source=fedsrch |