|本期目录/Table of Contents|

[1]吴冬,张野,薛伯遒,等.基于注意力增强型残差网络的火灾烟雾体积分数检测研究*[J].中国安全生产科学技术,2026,22(2):154-163.[doi:10.11731/j.issn.1673-193x.2026.02.019]
　WU Dong,ZHANG Ye,XUE Boqiu,et al.Research on fire smoke volume fraction detection based on attention-enhanced residual network[J].Journal of Safety Science and Technology,2026,22(2):154-163.[doi:10.11731/j.issn.1673-193x.2026.02.019]
点击复制

基于注意力增强型残差网络的火灾烟雾体积分数检测研究*()

《中国安全生产科学技术》[ISSN:1673-193X/CN:11-5335/TB]

卷:
22
期数:
2026年2期
页码:
154-163
栏目:
火灾与爆炸安全
出版日期:
2026-02-28

文章信息/Info

Title:
Research on fire smoke volume fraction detection based on attention-enhanced residual network
文章编号:
1673-193X(2026)-02-0154-10
作者:
吴冬; 张野; 薛伯遒; 李畅; 阎卫东
（1.沈阳建筑大学 土木工程学院，辽宁 沈阳 110168；
2.铁岭市消防救援支队，辽宁 铁岭 112008)
Author(s):
WU Dong;  ZHANG Ye;  XUE Boqiu;  LI Chang;  YAN Weidong
（1.School of Civil Engineering,Shenyang Jianzhu University,Shenyang Liaoning 110168,China;
2.Tieling Fire and Rescue Detachment,Tieling Liaoning 112008,China)
关键词:
烟雾体积分数; 注意力增强型残差网络; 图像特征
Keywords:
smoke volume fraction;  attention-enhanced residual network;  image features
分类号:
X932
DOI:
10.11731/j.issn.1673-193x.2026.02.019
文献标志码:
A
摘要:
为解决传统接触式烟雾传感器在火灾监测中存在的检测范围受限、精度低等问题，提出基于注意力增强型残差网络的烟雾体积分数预测方法。首先，采集多光照、多燃烧物场景下的烟雾图像与对应烟雾体积分数数据，构建“烟雾图像-体积分数”关联数据集（SIVF数据集）；其次，从颜色、亮度、纹理等7个维度提取27项烟雾图像特征变化量；最后，构建基于注意力增强型残差网络的火灾烟雾体积分数预测模型并对模型性能进行评估。研究结果表明：该模型在验证集上均方根误差RMSE为0.008 5%，平均绝对误差MAE为0.005 2%，决定系数R2达0.943，显著优于XGBoost、ResNet50、DCCNN等方法；在场景测试中，烟雾体积分数预测值与真实值趋势一致性较高，平均预测偏差为12.4%；模型推理速度达到21.05帧/s，满足实时预测需求。研究结果可为火灾早期预警、疏散路径规划等提供高精度、非接触式的烟雾体积分数量化方法。
Abstract:
In order to address the limitations of traditional contact-based smoke sensors,such as restricted detection range and low accuracy in fire monitoring,a smoke volume fraction prediction method based on an attention-enhanced residual network is proposed.Firstly,smoke images and corresponding smoke volume fraction data are collected under various lighting and combustion scenarios,creating the "Smoke Image-Volume Fraction" correlation dataset (SIVF dataset).Secondly,27 smoke image feature variations are extracted from seven dimensions,including color,brightness,and texture.Finally,a fire smoke volume fraction prediction model based on the attention-enhanced residual network is constructed,and the model’s performance is evaluated.The results show that the model achieves an RMSE of 0.0085%,an MAE of 0.0052%,and an R2 of 0.943 on the validation set,significantly outperforming methods such as XGBoost,ResNet50,and DCCNN.In scene testing,the predicted smoke volume fraction values show high trend consistency with the true values,with an average prediction deviation of 12.4%.The model’s inference speed reaches 21.05 frames/s,meeting the real-time prediction requirements.The findings provide a high-precision,non-contact smoke volume fraction quantification method for early fire warning,evacuation path planning,and other applications.

参考文献/References:

［1］刘涛,王卫星.一种基于BIM的计算机辅助建筑物火灾风险态势评估方法［J］.消防科学与技术,2024,43(2):244-248.LIU Tao,WANG Weixing.A computer aided building fire risk assessment method based on BlM［J］.Fire Science and Technology,2024,43(2):244-248.
［2］陈嫣君,霍一诺,杨承林,等.基于空间感知与显著性建模的图像烟雾检测方法［J］.清华大学学报(自然科学版),2025,65(11):2168-2179.CHEN Yanjun,HUO Yinuo,YANG Chenglin,et al.An image-based smoke detection method integrating spatial perception and saliency modeling［J］.Journal of Tsinghua University(Science and Technology),2025,65(11):2168-2179.
［3］吴健,王志鹏,谢丹,等.变电站火灾监测预警及自动灭火智慧系统设计［J］.中国安全生产科学技术,2025,21(1):175-185.WU Jian,WANG Zhipeng,XIE Dan,et al.Design of intelligent system for fire monitoring and early-warning and automatic fire extinguishing of substation［J］.Journal of Safety Science and Technology,2025,21(1):175-185.
［4］HOU F F,RUI X Y,CHEN Y H,et al.Flame and smoke semantic dataset:indoor fire detection with deep semantic segmentation model［J］.Electronics,2023,12:3778.
［5］张明振,段江忠,梁肇伟,等.基于改进YOLO-V5算法的烟火检测方法［J］.中国安全科学学报,2024,34(5):155-161.ZHANG Mingzhen,DUAN Jiangzhong,LIANG Zhaowei,et al.Firework detection method based on improved YOLO-V5 algorithm［J］.China Safety Science Journal,2024,34(5):155-161.
［6］SAYDIRASULOVICH S N,MUKHIDDINOV M,DJURAEV O,et al.An improved wildfire smoke detection based on YOLOv8 and UAV images［J］.Sensors,2023,23:8374.
［7］刘邵南,阮晨冉,陈建萍,等.舰船典型火灾的火焰探测识别方法研究［J］.中国安全生产科学技术,2025,21(9):5-12.LlU Shaonan,RUAN Chenran,CHEN Jianping,et al.Study on flame detection and identification methods for typical ship fires［J］.Journal of Safety Science and Technology,2025,21(9):5-12.
［8］SUN Y,JIANG L,PAN J,et al.A satellite imagery smoke detection framework based on the Mahalanobis distance for early fire identification and positioning［J］.International Journal of Applied Earth Observation and Geoinformation,2023,118:103257.
［9］贾智瑞.基于彩色图像特征相似性的烟雾浓度视频测量方法研究［D］.太原:太原理工大学,2023.
［10］莫鸿飞,谢振平.基于双通道深度CNN的烟雾浓度测量方法［J］.模式识别与人工智能,2021,34(9):844-852.MO Hongfei,XIE Zhenping.Smoke density measurement method based on dual-channel deep CNN［J］.Pattern Recognition and Artificial Intelligence,2021,34(9):844-852.
［11］吴则琪,王坤,刘鵾鹏,等.基于视频图像的建筑内走道烟雾识别定位［J］.科学技术与工程,2024,24(22):9681-9688.WU Zeqi,WANG Kun,LlU Kunpeng,et al.Smoke recognition and localization in building corridors based on video images［J］.Science Technology and Engineering,2024,24(22):9681-9688.
［12］PREMA C E,SURESH S,KRISHNAN M N,et al.A novel efficient video smoke detection algorithm using co-occurrence of local binary pattern variants［J］.Fire Technology,2022,58,3139-3165.
［13］陈贵亮,金天宇,杨冬.动态场景下火灾烟雾多特征融合识别定位研究［J］.传感器与微系统,2025,44(6):33-37.CHEN Guiliang,JIN Tianyu,YANG Dong.Research on multi-feature fusion for fire smoke recognition and localization in dynamic scenarios［J］.Transducer and Microsystem Technologies,2025,44(6):33-37.
［14］刘通,程江华,华宏虎,等.结合YdUaVa颜色模型和改进MobileNetV3的视频烟雾检测方法［J］.国防科技大学学报,2021,43(5):80-85.LIU Tong,CHENG Jianghua,HUA Honghu,et al.Video smoke detection method combining YdUaVa color model and improved MobileNetV3［J］.Journal of National University of Defense Technology,2021,43(5):80-85.
［15］曹绪民,刘春晓,张金栋,等.基于亮度对比度增强与饱和度补偿的快速图像去雾算法［J］.计算机辅助设计与图形学学报,2018,30(10):1925-1934.CAO Xumin,LIU Chunxiao,ZHANG Jindong,et al.Fast image defogging algorithm based on luminance contrast enhancement and saturation compensation［J］.Journal of Computer-Aided Design & Computer Graphics,2018,30(10):1925-1934.
［16］杨波,张立娜,韩霄松.基于LBP和GLCM的肠道肿瘤图像特征提取方法［J］.吉林大学学报(理学版),2022,60(4):911-918.YANG Bo,ZHANG Li’na,HAN Xiaosong.Feature extraction method of intestinal tumor lmages based on LBP and GLCM［J］.Journal of Jilin University(Science Edition),2022,60(4):911-918.
［17］邓实强,丁浩,杨孟,等.基于视频图像的公路隧道火灾烟雾检测［J］.隧道建设(中英文),2022,42(2):291-302.DENG Shiqiang,DING Hao,YANG Meng,et al.Fire smoke detection in highway tunnels based on video lmages［J］.Tunnel Construction,2022,42(2):291-302.
［18］XIE Y,SU H,LI T,et al.Smoke diffusion feature extraction and concentration prediction method based on canny edge detection［C］//2024 14th International Symposium on Antennas,Propagation and EM Theory (ISAPE),Hefei,China,2024,1-4.
［19］WEN G,ZHOU F,MA Y,et al.A dense multi-scale context and asymmetric pooling embedding network for smoke segmentation［J］.IET Computer Vision,2024,18(2):236-246.
［20］CHENG G,YANG L,YU Z,et al.A hybrid attention framework integrating channel-spatial refinement and frequency spectral analysis for remote sensing smoke recognition［J］.Fire,2025,8:197.
［21］KIKUTA K,MURATA K.T,MURAKAMI Y.A daytime smoke detection method based on variances of optical flow and characteristics of HSV color on footage from outdoor camera in urban city［J］.Fire Technology,2024,60:1427-1452.
［22］滕诗宇,何丽君.融合多特征与全局-局部Transformer的图像修复算法［J］.电子测量技术,2025,48(6):121-129.TENG Shiyu,HE Lijun.Fusion of multi-features and global-local Transformer for image inpainting［J］.Electronic Measurement Technology,2025,48(6):121-129.

相似文献/References:

备注/Memo

备注/Memo:
收稿日期： 2025-10-21
* 基金项目： 辽宁省科学技术厅项目（2021JH2/10100005）
作者简介： 吴冬，博士研究生，主要研究方向为火灾场景人员疏散、图像识别与处理。
更新日期/Last Update: 2026-03-09