基于EE-YOLOv8s的多场景火灾迹象检测算法

doi:10.11996/JG.j.2095-302X.2025010013

摘要/Abstract

摘要：

针对目前烟火场景检测中，光照变化、烟火动态性、复杂背景、目标过小等干扰因素导致的火灾迹象目标误检和漏检的问题，提出一种YOLOv8s改进模型EE-YOLOv8s。设计MBConv-Block卷积模块融入YOLOv8的Backbone部分，实现EfficientNetEasy特征提取网络，保证模型轻量化的同时，优化图像特征提取；引入大型可分离核注意力机制LSKA改进SPPELAN模块，将空间金字塔部分改进为SPP_LSKA_ELAN，充分捕获大范围内的空间细节信息，在复杂多变的火灾场景中提取更全面的特征，从而区分目标与相似物体的差异；Neck部分引入可变形卷积DCN和跨空间高效多尺度注意力EMA，实现C2f_DCN_EMA可变形卷积校准模块，增强对烟火目标边缘轮廓变化的适应能力，促进特征的融合与校准，突出目标特征；在Head部分增设携带有轻量级、无参注意力机制SimAM的小目标检测头，并重新规划检测头通道数，加强多尺寸目标表征能力的同时，降低冗余以提高参数有效利用率。实验结果表明，改进后的EE-YOLOv8s网络模型相较于原模型，其参数量减少了13.6%，准确率提升了6.8%，召回率提升了7.3%，mAP提升了5.4%，保证检测速度的同时，提升了火灾迹象目标的检测性能。

关键词: 烟火目标检测, EfficientNetEasy主干网络, 大型可分离核注意力机制, 可变形卷积校准模块, 小目标检测

Abstract:

To mitigate the current issues of spurious and missed detections of fire signs in smoke and fire scene detection, caused by interfering factors such as illumination variations, fire dynamics, complex backgrounds, and excessively small targets, an improved YOLOv8s model named EE-YOLOv8s was proposed. The EE-YOLOv8s model integrated the MBConv-Block convolution module into the YOLOv8 Backbone and employed the EfficientNetEasy feature extraction network to refine image feature extraction while preserving a lightweight design. Additionally, the SPPELAN module was upgraded to SPP_LSKA_ELAN by incorporating the large separable kernel attention mechanism (LSKA), which captured spatial detail information in intricate and dynamic fire scenes, thereby distinguishing target objects from convoluted backgrounds. The Neck section introduced deformable convolution (DCN) and cross-space efficient multi-scale attention (EMA), implementing the C2f_DCN_EMA deformable convolution calibration module to enhance the adaptation to edge contour changes of fire and smoke targets, facilitating feature fusion and calibration, and emphasizing key target features. A small target detection head, equipped with the lightweight, parameter-free attention mechanism SimAM, was integrated into the Head section, and the channel configuration was refined to strengthen multi-size target characterization while minimizing redundancy and maximizing parameter utilization efficiency. Experimental results demonstrated that EE-YOLOv8s reduced the parameter count by 13.6%, while improving accuracy by 6.8%, recall by 7.3%, and mAP by 5.4% compared to the original model, ensuring rapid detection speed and superior detection performance for fire targets.

Key words: smoke and fire target detection, EfficientNetEasy backbone network, large separable kernel attention mechanism, deformable convolutional calibration module, small target detection

中图分类号:

TP391

崔克彬, 耿佳昌. 基于EE-YOLOv8s的多场景火灾迹象检测算法[J]. 图学学报, 2025, 46(1): 13-27.

CUI Kebin, GENG Jiachang. A multi-scene fire sign detection algorithm based on EE-YOLOv8s[J]. Journal of Graphics, 2025, 46(1): 13-27.

图/表 20

参考文献 32

[1]	PARK S, HAN K W, LEE K. A study on fire detection technology through spectrum analysis of smoke particles[C]// 2020 International Conference on Information and Communication Technology Convergence. New York: IEEE Press, 2020: 1563-1565.
[2]	XAVIER K L B L, NANAYAKKARA V K. Development of an early fire detection technique using a passive infrared sensor and deep neural networks[J]. Fire Technology, 2022, 58(6): 3529-3552.
[3]	KIM Y H, KIM A, JEONG H Y. RGB color model based the fire detection algorithm in video sequences on wireless sensor network[J]. International Journal of Distributed Sensor Networks, 2014, 10(4): 1-10.
[4]	肖潇, 孔凡芝, 刘金华. 基于动静态特征的监控视频火灾检测算法[J]. 计算机科学, 2019, 46(S1): 284-286, 299.
	XIAO X, KONG F Z, LIU J H. Monitoring video fire detection algorithm based on dynamic characteristics and static characteristics[J]. Computer Science, 2019, 46(S1): 284-286, 299 (in Chinese).
[5]	HASHEMZADEH M, ZADEMEHDI A. Fire detection for video surveillance applications using ICA K-medoids-based color model and efficient spatio-temporal visual features[J]. Expert Systems with Applications, 2019, 130: 60-78.
[6]	刘树东, 姚文渤, 张艳. 雾天条件下基于机器视觉的森林火灾监测[J]. 计算机工程与科学, 2020, 42(7): 1253-1261.
	LIU S D, YAO W B, ZHANG Y. Forest fire monitoring based on machine vision in foggy weather[J]. Computer Engineering & Science, 2020, 42(7): 1253-1261 (in Chinese).
[7]	PAN J, OU X M, XU L. A collaborative region detection and grading framework for forest fire smoke using weakly supervised fine segmentation and lightweight faster-RCNN[J]. Forests, 2021, 12(6): 768.
[8]	皮骏, 刘宇恒, 李久昊. 基于YOLOv5s的轻量化森林火灾检测算法研究[J]. 图学学报, 2023, 44(1): 26-32. DOI
	PI J, LIU Y H, LI J H. Research on lightweight forest fire detection algorithm based on YOLOv5s[J]. Journal of Graphics, 2023, 44(1): 26-32 (in Chinese). DOI
[9]	赵辉, 赵尧, 金林林, 等. 基于YOLOX的小目标烟火检测技术研究与实现[J]. 图学学报, 2022, 43(5): 783-790.
	ZHAO H, ZHAO Y, JIN L L, et al. Research and realization of small target smoke and fire detection technology based on YOLOX[J]. Journal of Graphics, 2022, 43(5): 783-790 (in Chinese).
[10]	LI A, ZHAO Y Q, ZHENG Z X. Novel recursive BiFPN combining with swin transformer for wildland fire smoke detection[J]. Forests, 2022, 13(12): 2032.
[11]	KIM S Y, MUMINOV A. Forest fire smoke detection based on deep learning approaches and unmanned aerial vehicle images[J]. Sensors, 2023, 23(12): 5702.
[12]	LI C D, LI G Y, SONG Y C, et al. Fast forest fire detection and segmentation application for UAV-assisted mobile edge computing system[J]. IEEE Internet of Things Journal, 2024, 11(16): 26690-26699.
[13]	HUANG J W, ZHOU J S, YANG H Z, et al. A small-target forest fire smoke detection model based on deformable transformer for end-to-end object detection[J]. Forests, 2023, 14(1): 162.
[14]	JIN L, YU Y Q, ZHOU J N, et al. SWVR: a lightweight deep learning algorithm for forest fire detection and recognition[J]. Forests, 2024, 15(1): 204.
[15]	XU Y Q, LI J M, ZHANG L, et al. CNTCB-YOLOv7: an effective forest fire detection model based on ConvNeXtV2 and CBAM[J]. Fire, 2024, 7(2): 54.
[16]	DILSHAD N, KHAN S U, ALGHAMDI N S, et al. Toward efficient fire detection in IoT environment: a modified attention network and large-scale data set[J]. IEEE Internet of Things Journal, 2024, 11(8): 13467-13481.
[17]	WANG C Y, BOCHKOVSKIY A, LIAO H Y M. YOLOv7: trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 7464-7475.
[18]	TAN M X, LE Q V. EfficientNetV2: smaller models and faster training[EB/OL]. [2024-05-10]. https://arxiv.org/abs/2104.00298.
[19]	LIU X Y, PENG H W, ZHENG N X, et al. EfficientViT: memory efficient vision transformer with cascaded group attention[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 14420-14430.
[20]	TAN M X, LE Q V. EfficientNet: rethinking model scaling for convolutional neural networks[EB/OL]. [2024-05-10]. https://arxiv.org/abs/1905.11946.
[21]	WANG A, CHEN H, LIN Z J, et al. RepVIT: revisiting mobile CNN from ViT perspective[EB/OL]. [2024-05-10]. https://arxiv.org/abs/2307.09283.
[22]	WANG W H, DAI J F, CHEN Z, et al. InternImage: exploring large-scale vision foundation models with deformable convolutions[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 14408-14419.
[23]	OUYANG D L, HE S, ZHANG G Z, et al. Efficient multi-scale attention module with cross-spatial learning[C]// 2023 IEEE International Conference on Acoustics, Speech and Signal Processing. New York: IEEE Press, 2023: 1-5.
[24]	WANG C Y, YEH I H, LIAO H Y M. YOLOv9: learning what you want to learn using programmable gradient information[EB/OL]. [2024-04-10]. https://arxiv.org/abs/2402.13616.
[25]	LAU K W, PO L M, REHMAN Y A U. Large separable kernel attention: rethinking the large kernel attention design in CNN[J]. Expert Systems with Applications, 2024, 236: 121352.
[26]	YANG L X, ZHANG R Y, LI L D, et al. SimAM: a simple, parameter-free attention module for convolutional neural networks[EB/OL]. [2024-04-05]. https://proceedings.mlr.press/v139/yang21o.html.
[27]	WANG M, JIANG L C, YUE P, et al. FASDD: an open-access 100,000-level flame and smoke detection dataset for deep learning in fire detection[EB/OL]. [2024-04-05]. https://doi. org/10.5194/essd-2023-73.
[28]	CHEN J R, KAO S H, HE H, et al. Run, don't walk: chasing higher FLOPS for faster neural networks[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 12021-12031.
[29]	LI C Y, LI L L, JIANG H L, et al. YOLOv6: a single-stage object detection framework for industrial applications[EB/OL]. [2024-04-19]. https://arxiv.org/abs/2209.02976.
[30]	CARION N, MASSA F, SYNNAEVE G, et al. End-to-end object detection with transformers[C]// The 16th European Conference on Computer Vision. Cham: Springer, 2020: 213-229.
[31]	ZHAO Y, LV W Y, XU S L, et al. DETRs beat YOLOs on real-time object detection[EB/OL]. [2024-04-19]. https://arxiv.org/abs/2304.08069.
[32]	WENG K H, CHU X X, XU X M, et al. EfficientRep: an efficient repvgg-style ConvNets with hardware-aware neural network design[EB/OL]. [2024-04-19]. https://arxiv.org/abs/2302.00386.

名称	版本号
Operating System	Windows 11系统
CPU	12th Gen Intel(R) Core(TM) i5-12400F @2.50 GHz
显卡(GPU)	NVIDIA GeForce RTX 4060 Ti (16380 MiB)
Pytorch	torch - 1.13.1 + cu117
GPU accleration CUDA	CUDA 11.7 + cuDNN v8.9.6
Programming Language	Python - 3.9.19

名称	版本号
Operating System	Windows 11系统
CPU	12th Gen Intel(R) Core(TM) i5-12400F @2.50 GHz
显卡(GPU)	NVIDIA GeForce RTX 4060 Ti (16380 MiB)
Pytorch	torch - 1.13.1 + cu117
GPU accleration CUDA	CUDA 11.7 + cuDNN v8.9.6
Programming Language	Python - 3.9.19

Group	主干网络	P/%	R/%	mAP0.50/%	mAP0.75/%	mAP/%	Param/10⁶	GFLOPs	FPS
①	CSPDarkNets	83.6	80.2	89.5	71.8	62.9	11.14	28.6	120.78
②	EfficientNetv2s	85.7	84.0	88.4	72.4	66.3	18.03	40.7	44.73
③	EfficientViT M4	83.1	75.5	85.7	66.6	58.3	10.45	26.4	83.68
④	RepViT M 1.1	85.6	84.3	90.9	62.3	58.9	9.93	28.1	74.37
⑤	FasterNet T1	82.3	76.0	86.1	67.7	59.3	9.40	25.5	106.63
⑥	EfficientNetEasy	85.2	87.8	89.6	67.3	60.0	11.07	27.9	114.96

Group	主干网络	P/%	R/%	mAP0.50/%	mAP0.75/%	mAP/%	Param/10⁶	GFLOPs	FPS
①	CSPDarkNets	83.6	80.2	89.5	71.8	62.9	11.14	28.6	120.78
②	EfficientNetv2s	85.7	84.0	88.4	72.4	66.3	18.03	40.7	44.73
③	EfficientViT M4	83.1	75.5	85.7	66.6	58.3	10.45	26.4	83.68
④	RepViT M 1.1	85.6	84.3	90.9	62.3	58.9	9.93	28.1	74.37
⑤	FasterNet T1	82.3	76.0	86.1	67.7	59.3	9.40	25.5	106.63
⑥	EfficientNetEasy	85.2	87.8	89.6	67.3	60.0	11.07	27.9	114.96

Group	空间金字塔池化模块	P%	R%	mAP/%	Param/10⁶	GFLOPs	FPS
①	SPPF	83.6	80.2	62.9	11.14	28.6	120.78
②	SPPCSPC^[29]	83.6	78.9	61.7	17.55	33.6	124.97
③	SPPFCSPC	83.3	80.5	63.6	17.55	33.6	116.72
④	SPPF_LSKA	88.5	80.4	64.6	12.20	29.3	122.80
⑤	SPPELAN	83.1	84.6	64.4	10.67	27.9	117.12
⑥	SPP_LSKA_ELAN	88.3	85.1	66.4	12.45	29.3	115.46