Accurate and timely detection of incipient forest fire smoke is essential for forest fire prevention and ecosystem protection. However, existing image-based methods are often limited by irregular smoke appearance, complex forest backgrounds, and limited computational capacity on edge devices. This study proposes LCA-YOLO, a lightweight detection framework for edge-based forest fire smoke monitoring. The model incorporates Linear Deformable Convolution (LDConv) to improve adaptability to irregular smoke morphology, Content-Guided Attention Fusion (CGA-Fusion) to enhance feature representation under cluttered backgrounds, and Adaptive Threshold Focal Loss (ATFL) to increase the optimization emphasis on low-confidence hard samples, which may include faint early-stage smoke. Experimental results on a self-built forest fire smoke dataset and multiple public benchmarks demonstrated a favorable balance between detection accuracy and computational efficiency. With 2.39 M parameters, LCA-YOLO achieved an mAP@0.50 of 0.887 in the seed-0 run and an average mAP@0.50 of 0.8844 ± 0.0052 across five independent runs. Evaluation on an NVIDIA Jetson Nano at an input resolution of 480 × 480 pixels yielded an average inference latency of 78.6 ms per image, corresponding to approximately 12.7 FPS and demonstrating the feasibility of continuous edge-side smoke inference under the tested hardware configuration. These results support the computational feasibility of deploying LCA-YOLO on resource-constrained edge hardware.
| [1] |
Adarsh P, Rathi P, Kumar M (2020) YOLOv3-Tiny: object detection and recognition using one stage improved model. In: 2020 6th international conference on advanced computing and communication systems (ICACCS), Coimbatore, India, pp 687–694. https://doi.org/10.1109/ICACCS48705.2020.9074315
|
| [2] |
Aiformankind (2020) Open Wildfire Smoke Datasets. https://github.com/aiformankind/wildfire-smoke-dataset. Accessed 13 August 2026
|
| [3] |
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
|
| [4] |
Chaturvedi S, Thakur PS, Khanna P, Ojha A, Song Y, Awange JL. Satellite image-based surveillance and early wildfire smoke detection using a multiattention interlaced network. IEEE Trans Ind Inform, 2025, 21(53806-3815
|
| [5] |
Chen Y, Li J, Sun K, Zhang Y. A lightweight early forest fire and smoke detection method. J Supercomput, 2024, 80(7): 9870-9893
|
| [6] |
Chen Z, He Z, Lu ZM. DEA-Net: single image dehazing based on detail-enhanced convolution and content-guided attention. IEEE Trans Image Process, 2024, 33: 1002-1015
|
| [7] |
De Rango A, Furnari L, Cortale F, Senatore A, Mendicino G. Wildfire early warning system based on a smart CO2 sensors network. Sensors (Basel), 2025, 25(7 2012
|
| [8] |
Dewangan A, Pande Y, Braun HW, Vernon F, Perez I, Altintas I, Cottrell GW, Nguyen MH. FIgLib & SmokeyNet: dataset and deep learning model for real-time wildland fire smoke detection. Remote Sens, 2022, 14(4): 1007
|
| [9] |
Dollár P, Wojek C, Schiele B, Perona P. Pedestrian detection: an evaluation of the state of the art. IEEE Trans Pattern Anal Mach Intell, 2012, 34(4): 743-761
|
| [10] |
Everingham M, Van Gool L, Williams CKI, Winn J, Zisserman A. The PASCAL visual object classes (VOC) challenge. Int J Comput Vis, 2010, 88(2): 303-338
|
| [11] |
Ge Z, Liu S, Wang F, Li Z, Sun J (2021) YOLOX: exceeding YOLO series in 2021. arXiv preprint arXiv:2107.08430. https://arxiv.org/abs/2107.08430
|
| [12] |
Honary R, Shelton J, Kavehpour P. A review of technologies for the early detection of wildfires. ASME Open J Eng, 2025, 4 040803
|
| [13] |
Khudayberdiev O, Zhang J, Abdullahi SM, Zhang S. Light-FireNet: an efficient lightweight network for fire detection in diverse environments. Multimed Tools Appl, 2022, 81(17): 24553-24572
|
| [14] |
Lambrou N, Kolden C, Loukaitou-Sideris A. Disaster recovery gentrification in post-wildfire landscapes: the case of Paradise. Int J Disaster Risk Reduct, 2025, 118 105235
|
| [15] |
Li L. A comprehensive survey of Fire Weather Index (FWI) systems and IoT applications in peatland fire management. IEEE Access, 2025, 13: 33579-33599
|
| [16] |
Li T, Zhu H, Hu C, Zhang J. An attention-based prototypical network for forest fire smoke few-shot detection. J For Res, 2022, 33(5): 1493-1504
|
| [17] |
Li C, Li L, Geng Y, Jiang H, Cheng M, Zhang B, Ke Z, Xu X, Chu X (2023) YOLOv6 V3.0: a full-scale reloading. arXiv preprint arXiv:2301.05586. https://arxiv.org/abs/2301.05586
|
| [18] |
Lin TY, Maire M, Belongie S, Hays J, Perona P, Ramanan D, Dollár P, Zitnick CLFleet D, Pajdla T, Schiele B, Tuytelaars T. Microsoft COCO: common objects in context. Computer vision—ECCV 2014: 13th European conference, Zurich, Switzerland, September 6–12, 2014, proceedings, part V, 2014, Cham, Springer, 740-755 vol 8693
|
| [19] |
Liu W, Anguelov D, Erhan D, Szegedy C, Reed S, Fu CY, Berg ACLeibe B, Matas J, Sebe N, Welling M. SSD: single shot MultiBox detector. Computer vision—ECCV 2016: 14th European conference, Amsterdam, The Netherlands, October 11–14, 2016, proceedings, Part I, 2016, Cham, Springer, pp 21-37 vol 9905
|
| [20] |
Liu Y, Chen F, Zhang C, Wang Y, Zhang J. Early wildfire smoke detection method based on EDA. Remote Sens, 2024, 16(24): 4684
|
| [21] |
Liu Z, Lin Y, Cao Y, Hu H, Wei Y, Zhang Z, Lin S, Guo B (2021) Swin Transformer: hierarchical vision transformer using shifted windows. In: 2021 IEEE/CVF international conference on computer vision (ICCV), Montreal, QC, Canada, pp 9992–10002. https://doi.org/10.1109/ICCV48922.2021.00986
|
| [22] |
Qiao L, Li S, Zhang Y, Yan J. Early wildfire detection and distance estimation using aerial visible-infrared images. IEEE Trans Ind Electron, 2024, 71(12): 16695-16705
|
| [23] |
Ramos L, Casas E, Bendek E, Romero C, Rivas-Echeverría F. Computer vision for wildfire detection: a critical brief review. Multimed Tools Appl, 2024, 83(35): 83427-83470
|
| [24] |
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
|
| [25] |
Rukundo O. Effects of image size on deep learning. Electronics, 2023, 12(4): 985
|
| [26] |
Sandler M, Howard A, Zhu M, Zhmoginov A, Chen LC (2018) MobileNetV2: inverted residuals and linear bottlenecks. In: 2018 IEEE/CVF conference on computer vision and pattern recognition (CVPR), Salt Lake City, UT, USA, pp 4510–4520. https://doi.org/10.1109/CVPR.2018.00474
|
| [27] |
Sokolova M, Lapalme G. A systematic analysis of performance measures for classification tasks. Inf Process Manag, 2009, 45(4): 427-437
|
| [28] |
Ultralytics (2020) Ultralytics YOLOv5. https://github.com/ultralytics/yolov5. Accessed 13 August 2026
|
| [29] |
Ultralytics (2023) Ultralytics YOLOv8. https://github.com/ultralytics/ultralytics. Accessed 13 August 2026
|
| [30] |
Ultralytics (2024) Ultralytics YOLO11 (Version 11.0.0). https://github.com/ultralytics/ultralytics. Accessed 13 August 2026
|
| [31] |
Wang A, Chen H, Liu L, Chen K, Lin Z, Han J, Ding G. YOLOv10: real-time end-to-end object detection. Adv Neural Inf Process Syst, 2024, 37: 107984-108011
|
| [32] |
Wang CY, Yeh IH, Liao HYMLeonardis A, Ricci E, Roth S, Russakovsky O, Sattler T, Varol G. YOLOv9: learning what you want to learn using programmable gradient information. Computer vision—ECCV 2024: 18th European conference, Milan, Italy, September 29–October 4, 2024, proceedings, Part XXXI, 2025, Cham, Springer, pp 1-21 vol 15089
|
| [33] |
Xiao Z, Wan F, Lei G, Xiong Y, Xu L, Ye Z, Liu W, Zhou W, Xu C. FL-YOLOv7: a lightweight small object detection algorithm in forest fire detection. Forests, 2023, 14(9): 1812
|
| [34] |
Xu G, Zhang Y, Zhang Q, Lin G, Wang J. Deep domain adaptation based video smoke detection using synthetic smoke images. Fire Saf J, 2017, 93: 53-59
|
| [35] |
Xu G, Zhang Q, Liu D, Lin G, Wang J, Zhang Y. Adversarial adaptation from synthesis to reality in fast detector for smoke detection. IEEE Access, 2019, 7: 29471-29483
|
| [36] |
Yang B, Zhang X, Zhang J, Luo J, Zhou M, Pi Y. EFLNet: enhancing feature learning network for infrared small target detection. IEEE Trans Geosci Remote Sens, 2024, 62: 1-11
|
| [37] |
Zhang Q, Lin G, Zhang Y, Xu G, Wang J. Wildland forest fire smoke detection based on Faster R-CNN using synthetic smoke images. Procedia Eng, 2018, 211: 441-446
|
| [38] |
Zhang X, Song Y, Song T, Yang D, Ye Y, Zhou J, Zhang L. LDConv: linear deformable convolution for improving convolutional neural networks. Image Vis Comput, 2024, 149 105190
|
| [39] |
Zhu P, Wen L, Du D, Bian X, Fan H, Hu Q, Ling H. Detection and tracking meet drones challenge. IEEE Trans Pattern Anal Mach Intell, 2022, 44(11): 7380-7399
|
Funding
Fundamental Research Funds for the Central Universities(QNTD202304)
Xiong’an New Area Science and Technology Innovation Special Project of Ministry of Science and Technology of China(2023XAGG0065)
Rights & permissions
Northeast Forestry University
PDF
