The method of smoke and flame target extraction facing UAV and remote sensing images

doi:10.13474/j.cnki.11-2246.2025.1202

Abstract

Abstract: Addressing the bottlenecks in UAV and remote sensing-based smoke/flame detection task,such as insufficient multi-scale feature capture,complex background interference,and blurred edges,a method of smoke and flame target extraction based on the improved YOLOv12 model is proposed in this paper.The proposed method enhances multi-scale feature fusion through a mixed local-channel attention (MLCA)mechanism,improves detail retention in low-resolution images via an adaptive downsampling (ADown)module,and refines boundary regression accuracy with a customized adaptive loss function.Furthermore,by integrating the fine-tuned SAM2.1 model,the paper's method can realize pixel-level segmentation of targets within detection boxes.Experiments on the FASDD＿UAV,FASDD＿RS,and S-Firedata datasets shows that the proposed method achieves mAP50 scores of 93.1%,78.5%,and 68.2%,outperforming the baseline model YOLOv12 by 1.3%,1.5%,and 1.2%,respectively.The proposed method demonstrates significant advantages in detecting small targets,handling occluded scenarios and complex lighting conditions.Additionally,ablation experiments have confirmed the feature enhancement effects of the MLCA and Adown modules,as well as the optimization effect of Focaler-PIoU on model performance through dynamic gradient allocation.

Key words: remote sensing images, smoke flame detection, YOLO, SAM, attention mechanism

CLC Number:

P237

LIU Xiaodong, ZHAO Chenmeng, REN Yinghua, YANG Liping, ZHAO Like, ZHANG Ka. The method of smoke and flame target extraction facing UAV and remote sensing images[J]. Bulletin of Surveying and Mapping, 2025, 0(12): 7-14.

References

[1] RAFFERTY J P.Los Angeles wildfires of 2025[EB/OL].[2025-03-01].https://www.britannica.com/event/Los-Angeles-wildfires-of-2025.
[2] 赵柔嘉, 姚顺雨, 庞兴航.内蒙古森林消防:转型强能全力守护北疆万里绿色长城[J].中国安全生产, 2024, 19(8):38-39.
[3] 方超颖, 许军, 丁志龙, 等.山火灾害下输电系统的弹性评估方法及其提升措施[J].电力工程技术, 2024, 43(2):239-247.
[4] 苗利明, 邱树伟, 张杰洵, 等.基于嵌入式的古建筑火灾超早期检测系统的实现[J].软件工程, 2023, 26(1):59-62.
[5] YANG Songxi, HUANG Qunying, YU Manzhu.Advancements in remote sensing for active fire detection: a review of datasets and methods[J].Science of The Total Environment, 2024, 943: 173273.
[6] 沈龙, 钱国超, 彭兆裕, 等.绝缘子污秽等级的高光谱特征优化识别技术研究[J].电力工程技术, 2022, 41(2):156-162.
[7] XIA Jingjing, ZHOU Yi, ZENG Jin.Dual-wavelength optical sensor for fire detection and measurement of aerosol mass concentration[J].Fire Safety Journal, 2024, 146: 104129.
[8] 马庆禄, 马恋, 孔国英, 等.基于红外热成像的隧道火焰检测技术研究[J].火灾科学, 2022, 31(4):244-251.
[9] YANDOUZI M, BERRAHAL M, GRARI M, et al.Semantic segmentation and thermal imaging for forest fires detection and monitoring by drones[J].Bulletin of Electrical Engineering and Informatics, 2024, 13(4): 2784-2796.
[10] 安金铭, 梁秀满, 王雁, 等.Real-ESRGAN网络与形态学结合的烧结断面火焰图像复原方法[J].华北理工大学学报(自然科学版), 2023, 45(2):60-68.
[11] 赵振强, 何水原, 梁永志.基于Faster R-CNN的遥感影像船舶检测识别[J].测绘通报, 2021(11):59-64.
[12] ZHANG Haoran, YIN Zhenyu, ZHANG Ning, et al.A scale-adaptive mask R-CNN strategy for foreground particle segmentation and geometrical analysis of granular aggregates[J].Applied Soft Computing, 2024, 164: 111931.
[13] SHAMSOSHOARA A, AFGHAH F, RAZI A, et al.Aerial imagery pile burn detection using deep learning: the FLAME dataset[J].Computer Networks, 2021, 193: 108001.
[14] 荆启文, 郝思鹏, 李思源.基于改进YOLOv8算法的10kV配电线路导线裸露检测[J].电力工程技术, 2025, 44(3):201-210.
[15] WANG C, HE W, NIE Y, et al.Gold-YOLO: efficient object detector via gather-and-distribute mechanism[EB/OL].[2025-03-01].https://doi.org/10.48550/arXiv.2309.11331.
[16] 陈飞宇, 张应迁, 吴嘉懿, 等.基于GOLD-YOLO改进YOLOv5模型道路病害检测研究[J].现代计算机, 2024, 30(19):7-12.
[17] AZAD R, AGHDAM E K, RAULAND A, et al.Medical image segmentation review: the success of U-Net[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024, 46(12): 10076-10095.
[18] 王靖超, 刘勇宏, 杨怡琳, 等.基于YOLOv5s的遥感卫星图像森林火灾检测算法[J].电脑知识与技术, 2023, 19(23):8-12.
[19] TIAN Y, YE Q, DOERMANN D.YOLOv12: attention-centric real-time object detectors[EB/OL].[2025-03-01].https://doi.org/10.48550/arXiv.2502.12524.
[20] WANG Ming, YUE Peng, JIANG Liangcun, et al.An open flame and smoke detection dataset for deep learning in remote sensing based fire detection[J].Geo-spatial Information Science, 2025, 28(2): 511-526.
[21] 赵晨萌, 张卡, 盛业华, 等.多场景烟雾火焰目标检测数据集(S-Firedata)[EB/OL].[2025-03-01].http://geodata.nnu.edu.cn.
[22] RAVI N, GABEUR V, HU Y T, et al.Sam2: segment anything in images and videos[EB/OL].[2025-03-01].https://doi.org/10.48550/arXiv.2408.00714.
[23] WAN Dahang, LU Rongsheng, SHEN Siyuan, et al.Mixed local channel attention for object detection[J].Engineering Applications of Artificial Intelligence, 2023, 123: 106442.
[24] 陈腾杰, 李永安, 张之好, 等.基于改进YOLOv8n+DeepSORT的带式输送机异物检测及计数方法[J].工矿自动化, 2024, 50(8):91-98.
[25] GAO Xuliang, SUN Yubin, XIAO Yao, et al.Adaptive random down-sampling data augmentation and area attention pooling for low resolution face recognition[J].Expert Systems with Applications, 2022, 209: 118275.
[26] 冯金秋, 燕芳, 杨阳, 等.MMF-YOLO晶圆模具表面微缺陷检测算法[J].计算机工程与应用, 2025, 61(15):132-143.
[27] ZHENG Zhaohui, WANG Ping, LIU Wei, et al.Distance-IoU loss: faster and better learning for bounding box regression[J].Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 12993-13000.
[28] ZHANG H, ZHANG S.Focaler-IoU: more focused intersection over union loss[EB/OL].[2025-03-01].https://doi.org/10.48550/arXiv.2401.10525.
[29] LIU Can, WANG Kaige, LI Qing, et al.Powerful-IoU: more straightforward and faster bounding box regression loss with a nonmonotonic focusing mechanism[J].Neural Networks, 2024, 170: 276-284.
[30] 项家灏, 徐慧英, 徐广平, 等.MECW-YOLO:基于改进YOLOv8的无人机视角小目标检测算法[J/OL].计算机工程与科学, 1-12[2025-08-29].https://link.cnki.net/urlid/43.1258.TP.20241225.1106.008.
[31] WANG Gang, CHEN Yanfei, AN Pei, et al.UAV-YOLOv8: a small-object-detection model based on improved YOLOv8 for UAV aerial photography scenarios[J].Sensors, 2023, 23(16): 7190.
[32] 曹继卫, 罗飞, 丁炜超.BS-YOLO:基于BSAM注意力机制和SCConv的小目标检测算法[J/OL].计算机工程, 1-10[2025-08-29].https://doi.org/10.19678/j.issn.1000-3428.0070159.
[33] KHANAM R, HUSSAIN M.YOLOv11: an overview of the key architectural enhancements[EB/OL].[2025-03-01].https://doi.org/10.48550/arXiv.2410.17725.
[34] ZHAO Chenmeng, ZHAO Like, ZHANG Ka, et al.Smoke and fire-you only look once: a lightweight deep learning model for video smoke and flame detection in natural scenes[J].Fire, 2025, 8(3): 104.