Fire detection model in confined space based on improved SBR-YOLOv11n

doi:10.16265/j.cnki.issn1003-3033.2026.09.1007

Abstract

Abstract:

Aiming at the problems of low flame brightness, small target and easy interference in confined space fires, an improved confined space fire detection model SBR-YOLOv11n was proposed based on YOLOv11n model. In the backbone network part, some standard convolution layers were replaced by Space to Depth Convolution (SPDConv) layers to solve the problem that the flame features were easy to lose under low brightness. In the neck network part, the Bidirectional Feature Pyramid Network (BiFPN) module was introduced to improve the fusion effect of small target flame features through bidirectional connections and learnable weight mechanisms. In addition, the Inverted Residual Mobile Block (iRMB) was used to adaptively weight the dynamic flame features to improve the robustness of the model to dynamic flame detection. Experimental results show that the SBR-YOLOv11n model has the best evaluation indexes. Compared with the YOLOv11n model, Precision (P), Recall (R), mean Average Precision (mAP@0.5) and mAP@0.5:0.95 are increase by 5.27 %, 3.81 %, 4.17 % and 3.03 % respectively, reaching 84.94 %, 83.83 %, 84.63 % and 62.97 %, with a detection speed of 75 Frames Per Second (FPS). The SBR-YOLOv11n model meets the needs of rapid and accurate fire detection in confined space and is suitable for fire detection in confined spaces.

Key words: confined space, fire detection, SPDConv-BiFPN-iRMB-YOLOv11n (SBR-YOLOv11n), feature extraction, attention mechanism

CLC Number:

X932爆炸安全与防火、防爆

Li Cong, Xu Wenbo, Teng Wenfei, Wang Chenhui, Wu Jiansong. Fire detection model in confined space based on improved SBR-YOLOv11n[J]. China Safety Science Journal, 2026, 36(9): 263-270.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Table 1

Table 2

Fig.10

References 18

[1]	刘全义, 朱博, 邓力, 等. 受限空间内小尺度油池火行为实验[J]. 科学技术与工程, 2021, 21(3): 1230-1235.
	Liu Quanyi, Zhu Bo, Deng Li, et al. Experiment on small scale pool fire behavior in confined space[J]. Science Technology and Engineering, 2021, 21(3): 1230-1235.
[2]	李青蔚, 朱烨瑞, 杨智翔, 等. 温度对受限空间锂电自行车燃烧特性的影响[J]. 中国安全科学学报, 2025, 35(3): 99-106. doi: 10.16265/j.cnki.issn1003-3033.2025.03.0620
	Li Qingwei, Zhu Yerui, Yang Zhixiang, et al. Influence of temperature on combustion features of lithium-ion electric bicycles in confined spaces[J]. China Safety Science Journal, 2025, 35(3): 99-106. doi: 10.16265/j.cnki.issn1003-3033.2025.03.0620
[3]	陈庆, 魏旭, 陈光, 等. 受限空间射流火形成和发展影响因素研究[J]. 中国安全科学学报, 2020, 30(7): 35-40. doi: 10.16265/j.cnki.issn1003-3033.2020.07.006
	Chen Qing, Wei Xu, Chen Guang, et al. Study on influence factors of formation and development of jet fire in confined space[J]. China Safety Science Journal, 2020, 30(7): 35-40. doi: 10.16265/j.cnki.issn1003-3033.2020.07.006
[4]	Shirwaikar R, Narvekar A, Hosamani A, et al. Real-time semi-occluded fire detection and evacuation route generation: leveraging instance segmentation for damage estimation[J]. Fire Safety Journal, 2025, 152: DOI:10.1016/j.firesaf.2025.104338.
[5]	Liu Chebin, Ahuja N. Vision based fire detection[C]. Proceedings of the 17^th International Conference on Pattern Recognition, 2004. ICPR 2004. IEEE, 2004: 134-137.
[6]	Foggia P, Saggese A, Vento M. Real-time fire detection for video-surveillance applications using a combination of experts based on color, shape, and motion[J]. IEEE Transactions on Circuits & Systems for Video Technology, 2015, 25(9): 1545-1556.
[7]	Li Tingting, Zhao Entao, Zhang Junguo, et al. Detection of wildfire smoke images based on a densely dilated convolutional network[J]. Electronics, 2019, 8(10): DOI:10.3390/electronics8101131.
[8]	高均益, 张伟, 李泽麟. YOLO-BFEPS: 一种高效注意力增强的跨尺度YOLOv10火灾检测模型[J]. 计算机科学, 2025, 52(增刊1): 424-432.
	Gao Junyi, Zhang Wei, Li Zelin. YOLO-BFEPS: Efficient attention-enhanced cross-scale YOLOv10 fire detection model[J]. Computer Science, 2025, 52(S1): 424-432.
[9]	Huo Yinuo, Zhang Qixing, Jia Yang, et al. A deep separable convolutional neural network for multiscale image-based smoke detection[J]. Fire Technology, 2022, 58(3): 1445-1468. doi: 10.1007/s10694-021-01199-7
[10]	Choi H S, Jeon M, Song K, et al. Semantic fire segmentation model based on convolutional neural network for outdoor image[J]. Fire Technology, 2021, 57(6): 3005-3019. doi: 10.1007/s10694-020-01080-z
[11]	Shahid M, Virtusio J J, Wu Yuhsien, et al. Spatio-temporal self-attention network for fire detection and segmentation in video surveillance[J]. IEEE Access, 2022, 10: 1259-1275. doi: 10.1109/ACCESS.2021.3132787
[12]	许亚晨, 施珮, 陆俊. 改进的YOLOv5s轻量级城市火灾检测算法[J]. 计算机工程与设计, 2025, 46(8): 2305-2311.
	Xu Yachen, Shi Pei, Lu Jun. Improved YOLOv5s-based algorithm for lightweight urban fire detection[J]. Computer Engineering and Design, 2025, 46(8): 2305-2311.
[13]	Dalal S, Lilhore U K, Radulescu M, et al. A hybrid LBP-CNN with YOLO-v5-based fire and smoke detection model in various environmental conditions for environmental sustainability in smart city[J]. Environmental Science and Pollution Research, 2024: DOI:10.1007/s11356-024-32023-8.
[14]	Yan Hailong, Huang Junjian, Zhou Zhiyan. Plastic bottle localization and ranging using improved YOLO-PB and binocular stereo vision[J]. Measurement, 2025: DOI:10.1016/j.measurement.2025.118469.
[15]	Sunkara R, Luo Tie. No more strided convolutions or pooling: a new CNN building block for low-resolution images and small objects[C]. Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Cham: Springer Nature Switzerland, 2022: 443-459.
[16]	Tan Mingxing, Pang Ruoming, Le Q V. Efficientdet: Scalable and efficient object detection[C]. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2020: 10 781-10 790.
[17]	Zhang Jiangning, Li Xiangtai, Li Jian, et al. Rethinking mobile block for efficient attention-based models[C]. 2023 IEEE/CVF International Conference on Computer Vision (ICCV). IEEE Computer Society, 2023: 1389-1400.
[18]	Buriboev A S, Abduvaitov A, Jeon H S. Integrating color and contour analysis with deep learning for robust fire and smoke detection[J]. Sensors, 2025, 25(7): DOI:10.3390/s25072044.

序号	YOLOv11n	SPDConv	BiFPN	iRMB	P/%	R/%	mAP@0.5/%	mAP@0.5: 0.95/%	FPS/(帧/s)
1	√	×	×	×	79.67	80.02	80.46	59.94	113
2	√	√	×	×	80.88	79.56	80.91	61.83	107
3	√	×	√	×	81.30	80.62	81.06	60.30	89
4	√	×	×	√	79.87	80.81	80.62	61.27	90
5	√	√	√	×	82.93	81.99	82.30	62.98	85
6	√	√	×	√	83.41	79.93	81.19	61.80	96
7	√	×	√	√	81.51	82.35	82.06	62.57	83
8	√	√	√	√	84.94	83.83	84.63	62.97	75

模型	P/%	R/%	mAP@ 0.5/%	mAP@0.5: 0.95/%	FPS/ (帧/s)
YOLOv5n	74.49	77.22	77.52	56.76	138
YOLOv8n	77.46	79.52	78.99	58.50	116
YOLOv10n	76.91	78.47	79.16	57.24	121
YOLOv11n	79.67	80.02	80.46	59.94	113
SBR- YOLOv11n	84.94	83.83	84.63	62.97	75