A scale-shared feature-reconstruction network for rotation-aware forest fire detection under constrained UAV computing resources

Abstract

This paper proposes LSRO-YOLO (Lightweight Scale-Shared Rotated Object Detector), a lightweight oriented object detection architecture for Unmanned Aerial Vehicle (UAV) forest fire monitoring under resource-constrained onboard computing and complex canopy backgrounds. It addresses two major challenges in UAV wildfire detection: scale variations caused by changes in flight altitudes, and the irregular shapes and arbitrary orientations of early-stage wildland fire spots from top-down views. Specifically, an Efficient Up-Convolution Block (EUCB) is introduced to improve feature reconstruction under strict computational constraints. A C2f with Dilated-wise Reparam Block (C2f-DRB) enlarges the effective receptive field for diffuse smoke, and a Lightweight Scale-Shared Oriented Bounding Box head (LSS-OBB) enables orientation-aware localization with parameter sharing and model compression. In-domain evaluations on three complementary benchmarks demonstrate that LSRO-YOLO achieves mAP@0.5 scores of 90.8% on UAV-simulated Wildfire dataset, 85.7% on fixed-surveillance D-Fire dataset, and 96.5% on multi-scene Fire-8 dataset. The proposed method also shows competitive performance compared with recent advanced methods such as AsymmetricFPN, YOLOGX, and SlimNeck-YOLOv8. In addition, cross-dataset evaluation further demonstrates that the model, when trained on D-Fire, retains partial detection capability on unseen domains (mAP@0.5 scores: 53.1% on Fire-8, 35.0% on Wildfire). Relative to the YOLOv8n-OBB baseline, LSRO-YOLO reduces the parameter count by 25.3% to 2.30 M and lowers computational complexity by 18.1% to 6.8 GFLOPs. Further comparisons with recent lightweight detectors, including YOLO11n, YOLO26n, and D-FINE-n, show that LSRO-YOLO achieves the lowest parameter count while maintaining competitive detection accuracy. These results indicate LSRO-YOLO provides an effective accuracy–efficiency trade-off and represents a promising candidate architecture for memory-constrained UAV forest fire monitoring, while hardware-level validation on embedded platforms remains an essential next step toward practical deployment.

Keywords

YOLOv8 / Oriented object detection / Lightweight detection / Forest fire monitoring / UAV imagery

References

[1] Abonia S (2024) YOLOv8-fire-and-smoke-detection. GitHub repository

[2] Agbeshie AA, Abugre S, Atta-Darkwa T, Awuah R. A review of the effects of forest fire on soil properties. J for Res, 2022, 33(5): 1419-1441

[3] Baltzer JL, Walker XJ, Veraverbeke S, Hessilt TD, Alfaro-Sanchez R, Van Gerrevink MJ, Mack MC, Ogden EL, Olsen R, Scholten RC, Turetsky MR. Overwintering fires can occur in both peatlands and upland forests with varying ecological impacts. Nat Ecol Evol, 2025, 9(4): 559-564

[4] Ban Y, Zhang P, Nascetti A, Bevington AR, Wulder MA. Near real-time wildfire progression monitoring with Sentinel-1 SAR time series and deep learning. Sci Rep, 2020, 10(1): 1322

[5] Bochkovskiy A, Wang C-Y, Liao H-YM (2020) YOLOv4: Optimal Speed and Accuracy of Object Detection

[6] Cao Y, Lei R. Progress and prospect of forest fire monitoring based on the multi-source remote sensing data. Natl Remote Sens Bull, 2024, 28(8): 1854-1869

[7] Carta F, Zidda C, Putzu M, Loru D, Anedda M, Giusto D. Advancements in forest fire prevention: a comprehensive survey. Sensors (Basel), 2023, 23(14): 6635

[8] Chan CC, Alvi SA, Zhou X, Durrani S, Wilson N, Yebra M. A survey on IoT ground sensing systems for early wildfire detection: technologies, challenges, and opportunities. IEEE Access, 2024, 12: 172785-172819

[9] Cheng X, Bian J, Kang Y, Xie X, Deng Y, Lu Q, Tang J, Shi Y, Zhao J. Parameter efficient asymmetric feature pyramid for early wildfire detection. Appl Sci, 2025, 15(22): 12086

[10] Cui L, Ma R, Lv P, Jiang X, Gao Z, Zhou B, Xu M. MDSSD: multi-scale deconvolutional single shot detector for small objects. Sci China Inf Sci, 2020, 63(2): 120113

[11] Dampage U, Bandaranayake L, Wanasinghe R, Kottahachchi K, Jayasanka B. Forest fire detection system using wireless sensor networks and machine learning. Sci Rep, 2022, 12(1): 46

[12] De Venâncio PVAB, Lisboa AC, Barbosa AV. An automatic fire detection system based on deep convolutional neural networks for low-power, resource-constrained devices. Neural Comput Appl, 2022, 34(18): 15349-15368

[13] Deng L, Zhou J, Liu Q. Fire and smoke detection algorithm based on improved YOLOv8. J Tsinghua Unive Sci Technol, 2025, 65(4): 681-689

[14] Ding J, Xue N, Long Y, Xia G-S, Lu Q. Learning RoI Transformer for Oriented Object Detection in Aerial Images. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, Long Beach, CA, USA, IEEE2844-2853

[15] Ding X, Zhang Y, Ge Y, Zhao S, Song L, Yue X, Shan Y. UniRepLKNet: A Universal Perception Large-Kernel ConvNet for Audio, Video, Point Cloud, Time-Series and Image Recognition. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, Seattle, WA, USA, IEEE55135524

[16] Ge Z, Liu S, Wang F, Li Z, Sun J (2021) YOLOX: Exceeding YOLO Series in 2021

[17] Giannakidou S, Radoglou-Grammatikis P, Lagkas T, Argyriou V, Goudos S, Markakis EK, Sarigiannidis P. Leveraging the power of internet of things and artificial intelligence in forest fire prevention, detection, and restoration: a comprehensive survey. Internet Things, 2024, 26: 101171

[18] Jocher G, Changyu L, Hogan A, Lijun Yu 于力军, Changyu98, Rai P, Sullivan T (2020) ultralytics/yolov5: Initial Release

[19] Jocher G, Chaurasia A, Qiu J (2023) Ultralytics YOLO

[20] Jocher G, Qiu J, Chaurasia A (2024) Ultralytics YOLO

[21] Jocher G, Qiu J, Chaurasia A (2026) Ultralytics YOLO

[22] Kang Y, Jang E, Im J, Kwon C. A deep learning model using geostationary satellite data for forest fire detection with reduced detection latency. Gisci Remote Sens, 2022, 59(1): 2019-2035

[23] Khan S, Khan A. FFireNet: deep learning based forest fire classification and detection in smart cities. Symmetry (Basel), 2022, 14(10): 2155

[24] Kobziar LN, Hiers JK, Belcher CM, Bond WJ, Enquist CA, Loudermilk EL, Miesel JR, O’Brien JJ, Pausas JG, Hood S, Keane R, Morgan P, Pingree MRA, Riley K, Safford H, Seijo F, Varner JM, Wall T, Watts AC. Principles of fire ecology. Fire Ecol, 2024, 20(1): 39

[25] Li Z, Mihaylova L, Yang L. A deep learning framework for autonomous flame detection. Neurocomputing, 2021, 448: 205-216

[26] Li C, Li L, Jiang H, Weng K, Geng Y, Li L, Ke Z, Li Q, Cheng M, Nie W, Li Y, Zhang B, Liang Y, Zhou L, Xu X, Chu X, Wei X, Wei X (2022) YOLOv6: a single-stage object detection framework for industrial applications

[27] Li C, Du Y, Zhang X, Wu P. YOLOGX: an improved forest fire detection algorithm based on YOLOv8. Front Environ Sci, 2025, 12: 1486212

[28] Lin T-Y, Goyal P, Girshick R, He K, Dollar P. Focal Loss for Dense Object Detection. 2017 IEEE International Conference on Computer Vision (ICCV), 2017, Venice, IEEE2999-3007

[29] Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu C-Y, Berg ACLeibe B, Matas J, Sebe N, Welling M. SSD: single shot multibox detector. Computer vision – ECCV 2016, 2016, Cham, Springer International Publishing2137

[30] Lv W, Zhao Y, Chang Q, Huang K, Wang G, Liu Y (2024) RT-DETRv2: Improved baseline with bag-of-freebies for real-time detection transformer

[31] Pedrini PJL (2022) D-Fire: an image dataset for fire and smoke detection

[32] Peng Y, Li H, Wu P, Zhang Y, Sun X, Wu F (2024) D-FINE: redefine regression task in DETRs as fine-grained distribution refinement

[33] Rahman MM, Munir M, Marculescu R. EMCAD: Efficient Multi-Scale Convolutional Attention Decoding for Medical Image Segmentation. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, Seattle, WA, USA, IEEE1176911779

[34] Redmon J, Divvala S, Girshick R, Farhadi A. You Only Look Once: Unified, Real-Time Object Detection. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, Las Vegas, NV, USA, IEEE779788

[35] Redmon J, Farhadi A (2018) YOLOv3: An Incremental Improvement

[36] Ren S, He K, Girshick R, Sun J. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Trans Pattern Anal Mach Intell, 2017, 39(6): 1137-1149

[37] Roboflow Universe (2024) Wildfire - v1 Dataset

[38] Scholten RC, Jandt R, Miller EA, Rogers BM, Veraverbeke S. Overwintering fires in boreal forests. Nature, 2021, 593(7859): 399-404

[39] Singh H, Ang L-M, Lewis T, Paudyal D, Acuna M, Srivastava PK, Srivastava SK. Trending and emerging prospects of physics-based and ML-based wildfire spread models: a comprehensive review. J for Res, 2024, 35(1): 135

[40] Su H, Wang X, Chen W, Ding N, Cui X, Bai M, Chen Z, Li M. A novel framework for identifying causes of forest fire events using environmental and temporal characteristics of the ignition point in fire footprint. Ecol Indic, 2024, 160: 111899

[41] Tan M, Pang R, Le QV. EfficientDet: Scalable and Efficient Object Detection. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, Seattle, WA, USA, IEEE10778-10787

[42] Wang C-Y, Liao H-YM, Wu Y-H, Chen P-Y, Hsieh J-W, Yeh I-H. CSPNet: A New Backbone that can Enhance Learning Capability of CNN. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2020, Seattle, WA, USA, IEEE1571-1580

[43] Wang S, Zhao J, Ta N, Zhao X, Xiao M, Wei H. A real-time deep learning forest fire monitoring algorithm based on an improved Pruned + KD model. J Real-Time Image Proc, 2021, 18(6): 2319-2329

[44] Wang C-Y, Bochkovskiy A, Liao H-YM (2022a) YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors

[45] Wang Y, Hua C, Ding W, Wu R. Real-time detection of flame and smoke using an improved YOLOv4 network. SIViP, 2022, 16(4): 1109-1116

[46] Wei H, Liu X, Xu S, Dai Z, Dai Y, Xu X (2023) DWRSeg: Rethinking Efficient Acquisition of Multi-scale Contextual Information for Real-time Semantic Segmentation

[47] Wu Y, He KFerrari V, Hebert M, Sminchisescu C, Weiss Y. Group normalization. Computer vision – ECCV 2018, 2018, Cham, Springer International Publishing319

[48] Yang H, Cai H, Yang G, Geng D, Sun L. Predicting the rate of spread of mixed-fuel surface fires in northeastern China using the Rothermel wildfire behaviour model: a laboratory study. J for Res, 2024, 35(1): 108

[49] Yu G, Chang Q, Lv W, Xu C, Cui C, Ji W, Dang Q, Deng K, Wang G, Du Y, Lai B, Liu Q, Hu X, Yu D, Ma Y (2021) PP-PicoDet: a Better Real-Time Object Detector on Mobile Devices

[50] Zhang X, Yu W, Siyuan W, Bangyong S. An improved lightweight fire detection algorithm based on cascade sparse query. Opto-Electron Eng, 2023, 50(10): 230216

[51] Zhao L, Zhi L, Zhao C, Zheng W. Fire-YOLO: a small target object detection method for fire inspection. Sustainability, 2022, 14(9): 4930

[52] Zhao J, Yue C, Wang J, Hantson S, Wang X, He B, Li G, Wang L, Zhao H, Luyssaert S. Forest fire size amplifies postfire land surface warming. Nature, 2024, 633(8031): 828-834

[53] Zhao Y, Lv W, Xu S, Wei J, Wang G, Dang Q, Liu Y, Chen J. DETRs Beat YOLOs on Real-time Object Detection. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, Seattle, WA, USA, IEEE1696516974

Funding

the Central Finance Forestry Science and Technology Promotion Demonstration Project "Promotion and Demonstration of Air-Ground Integrated Monitoring Technology for Forest and Grassland Surface Fire"(Grant No. Hei [2023]TG25)

Specialized Technical Service Project of Liangshui National Nature Reserve(Grant No. HFW250400009)

RIGHTS & PERMISSIONS

Northeast Forestry University