密集多尺度上下文和分层特征融合注意力的早期烟雾分割

doi:10.3778/j.issn.1002-8331.2411-0094

摘要/Abstract

摘要： 输电线路附近发生林火时高温、烟雾和飞灰等因素可能导致绝缘间隙迅速下降，引发输电线路跳闸。由于烟雾出现通常比火焰更早，因此准确的烟雾探测能为电网预留更多处置时间。针对早期烟雾目标小且具有半透明特性造成分割困难的问题，提出了基于密集多尺度上下文和分层特征融合注意力的早期烟雾分割网络。该网络包含密集多尺度上下文模块，通过空洞卷积并行处理以得到烟雾的上下文信息，然后将并行卷积串联以提升有效采样点数量。相较于空洞空间卷积池化金字塔模块，烟雾的信息利用率提升了8.07个百分点，有效采样点数量提升至313个。此外，引入了超分辨率亚像素重建模块，以改善插值法上采样造成的图像失真和信息丢失问题。网络中的分层特征融合注意力模块利用原始、低阶和高阶特征层之间的内在关系，以弥补小薄烟雾的高阶特征信息丢失。实验结果表明，该网络的平均交并比达到79.92%，比基线网络提高6.27个百分点，综合性能优于其他分割网络。可视化分割结果表明，该网络能有效提取小薄烟雾特征信息，边缘分割更加平滑，整体性能获得较大提升，对于输电线路林火的预防具有重要意义。

关键词: 输电线路林火, 早期烟雾分割, 密集多尺度上下文模块（DMC）, 超分辨率亚像素重建模块（SRR）, 分层特征融合注意力模块（HFA）

Abstract: When a wildfire occurs near power transmission lines, factors such as high temperatures, smoke, and flying ash can cause the insulation gap to rapidly decrease, leading to power line trips. Since smoke typically appears before flames, accurate smoke detection can provide the power grid with more time to respond. To address the challenge of early smoke segmentation, which is difficult due to the small size and semitransparent nature of the smoke, this paper proposes an early smoke segmentation network based on dense multi-scale context and hierarchical feature fusion attention. The network includes a dense multi-scale context module, which uses parallel atrous convolutions to capture the contextual information of smoke and then serially connects the parallel convolutions to increase the number of effective sampling points. Compared to the atrous spatial pyramid pooling (ASPP) module, this approach improves smoke information utilization by 8.07 percentage points and increases the number of effective sampling points to 313. Additionally, a super resolution subpixel reconstruction module is introduced to mitigate image distortion and information loss caused by interpolation upsampling. Finally, the hierarchical feature fusion attention module in the network leverages the intrinsic relationships between original, low-level, and high-level feature layers to compensate for the loss of high-level feature information in small thin smoke. Experimental results show that the proposed network achieves a mean intersection over union (mIoU) of 79.92%, which is 6.27 percentage points higher than the baseline network, and its overall performance surpasses other segmentation networks. Visualization of the segmentation results demonstrates that the proposed network effectively extracts small thin smoke features, achieves smoother boundary segmentation, and significantly enhances overall performance, making it highly significant for the prevention of wildfires near power transmission lines.

Key words: wildfire near transmission lines, early smoke segmentation, dense multi-scale context module (DMC), super resolution subpixel reconstruction module (SRR), hierarchical feature fusion attention module (HFA)

王开正, 谭义章, 付一桐, 曾瑶, 李露露. 密集多尺度上下文和分层特征融合注意力的早期烟雾分割[J]. 计算机工程与应用, 2026, 62(8): 241-254.

WANG Kaizheng, TAN Yizhang, FU Yitong, ZENG Yao, LI Lulu. Early Smoke Segmentation Based on Dense Multi-Scale Context and Hierarchical Feature Fusion Attention[J]. Computer Engineering and Applications, 2026, 62(8): 241-254.

参考文献

[1] 刘云鹏，刘一瑾，律方成，等.数字孪生技术在输变电设备中的应用前景与关键技术[J].高电压技术，2022，48（5）：1621-1633.
LIU Y P，LIU Y J，Lü F C，et al.Application prospect and key technology of digital twin in power transmission and transformation equipment[J].High Voltage Engineering，2022，48（5）：1621-1633.
[2] 展瑞琦，叶蕴霞，夏嘉航，等.半控和全控器件串联的耦合关断模型及协调配合策略[J].中国电机工程学报，2023，43（10）：3949-3961.
ZHAN R Q，YE Y X，XIA J H，et al.Coupled turn-off model and coordination strategy of semi-controlled and fully-controlled devices in series[J].Proceedings of the CSEE，2023，43（10）：3949-3961.
[3] 杨亚宇，邰能灵，谢卫，等.利用单端边界能量的直流输电线路全线速动保护[J].电工技术学报，2023，38（9）：2403-2417.
YANG Y Y，TAI N L，XIE W，et al.A whole-line fast protection scheme for HVDC transmission line based on single-ended boundary energy[J].Transactions of China Electrotechnical Society，2023，38（9）：2403-2417.
[4] 张旋，张玉敏，吉兴全，等.基于目标级联分析法的输电网结构优化[J].电力自动化设备，2023，43（2）：218-224.
ZHANG X，ZHANG Y M，JI X Q，et al.Optimal transmission switching based on analytical target cascading method[J].Electric Power Automation Equipment，2023，43（2）：218-224.
[5] 王开正，付一桐，钱国超，等.基于时空上下文的输电线路山火自动识别算法[J].电力系统保护与控制，2022，50（18）：83-91.
WANG K Z，FU Y T，QIAN G C，et al.An automatic identification algorithm for wildfire occurrences near transmission line corridors based on spatio-temporal context[J].Power System Protection and Control，2022，50（18）：83-91.
[6] 杨国林，蒋兴良，王茂政，等.输电线路单导线覆冰形状对直流大电流融冰时间的影响[J].电工技术学报，2024，39（9）：2916-2924.
YANG G L，JIANG X L，WANG M Z，et al.The impact of ice accumulation shape on the DC high current ice-melting time for a single conductor on power transmission line[J].Transactions of China Electrotechnical Society，2024，39（9）：2916-2924.
[7] 周游，隋三义，陈洁，等.基于Himawari-8静止气象卫星的输电线路山火监测与告警技术[J].高电压技术，2020，46（7）：2561-2569.
ZHOU Y，SUI S Y，CHEN J，et al.Monitor and alarm technology of wildfire occurrences in transmission lines corridors based on himawari-8 geostationary meteorological satellite[J].High Voltage Engineering，2020，46（7）：2561-2569.
[8] YUAN F N，ZHANG L，XIA X，et al.Deep smoke segmentation[J].Neurocomputing，2019，357：248-260.
[9] KHAN S，MUHAMMAD K，HUSSAIN T，et al.DeepSmoke：deep learning model for smoke detection and segmentation in outdoor environments[J].Expert Systems with Applications，2021，182：115125.
[10] WANG J J，ZHANG X M，JING K L，et al.Learning precise feature via self-attention and self-cooperation YOLOX for smoke detection[J].Expert Systems with Applications，2023，228：120330.
[11] YIN H，WEI Y R.An improved algorithm based on convolutional neural network for smoke detection[C]//Proceedings of the 2019 IEEE International Conferences on Ubiquitous Computing & Communications and Data Science and Computational Intelligence and Smart Computing，Networking and Services.Piscataway：IEEE，2019：207-211.
[12] LI C H，YANG B，DING H，et al.Real-time video-based smoke detection with high accuracy and efficiency[J].Fire Safety Journal，2020，117：103184.
[13] 吴健辉，何灿，何俊康，等.FSNet：基于频率特性的烟雾图像分割网络[J].控制理论与应用，2023，40（4）：702-712.
WU J H，HE C，HE J K，et al.FSNet：a network for smoke image segmentation based on frequency characteristics[J].Control Theory & Applications，2023，40（4）：702-712.
[14] 罗旭东，吴一全，陈金林.无人机航拍影像目标检测与语义分割的深度学习方法研究进展[J].航空学报，2024，45（6）：235-264.
LUO X D，WU Y Q，CHEN J L.Research progress on deep learning methods for object detection and semantic segmentation in UAV aerial images[J].Acta Aeronautica et Astronautica Sinica，2024，45（6）：235-264.
[15] 裴峻鹏，汪有崧，李增辉，等.注意力机制在视网膜血管分割中的应用综述[J].计算机工程与应用，2024，60（14）：50-65.
PEI J P，WANG Y S，LI Z H，et al.Comprehensive review on application of attention mechanism in retinal vessel segmentation[J].Computer Engineering and Applications，2024，60（14）：50-65.
[16] 谭卓昆，罗龙飞，王顺芳.基于融合特征与GA-SVM算法的脑疾病基因预测[J].计算机工程与应用，2023，59（5）：70-77.
TAN Z K，LUO L F，WANG S F.Prediction of disease gene based on fusion features and GA-SVM algorithm[J].Computer Engineering and Applications，2023，59（5）：70-77.
[17] LI Y B，CHAI G Q，WANG Y T，et al.ACE R-CNN：an attention complementary and edge detection-based instance segmentation algorithm for individual tree species identification using UAV RGB images and LiDAR data[J].Remote Sensing，2022，14（13）：3035.
[18] GAO C Q，WANG L，XIAO Y X，et al.Infrared small-dim target detection based on Markov random field guided noise modeling[J].Pattern Recognition，2018，76：463-475.
[19] ZHAO M X，CHENG L，YANG X，et al.TBC-Net：a real-time detector for infrared small target detection using semantic constraint[EB/OL].[2024-09-28].https：//arxiv.org/abs/2001.05852.
[20] ZHANG T F，LI L，CAO S Y，et al.Attention-guided pyramid context networks for detecting infrared small target under complex background[J].IEEE Transactions on Aerospace and Electronic Systems，2023，59（4）：4250-4261.
[21] LI B Y，XIAO C，WANG L G，et al.Dense nested attention network for infrared small target detection[J].IEEE Transactions on Image Processing，2023，32：1745-1758.
[22] HUANG L，DAI S S，HUANG T，et al.Infrared small target segmentation with multiscale feature representation[J].Infrared Physics & Technology，2021，116：103755.
[23] LI X Q，CHEN Z X，JONATHAN WU Q M，et al.3D parallel fully convolutional networks for real-time video wildfire smoke detection[J].IEEE Transactions on Circuits and Systems for Video Technology，2020，30（1）：89-103.
[24] YUAN F N，ZHANG L，XIA X，et al.A wave-shaped deep neural network for smoke density estimation[J].IEEE Transactions on Image Processing，2020，29：2301-2313.
[25] JING T，MENG Q H，HOU H R.SmokeSeger：a transformer-CNN coupled model for urban scene smoke segmentation[J].IEEE Transactions on Industrial Informatics，2024，20（2）：1385-1396.
[26] YUAN F N，DONG Z S，ZHANG L，et al.Cubic-cross convolutional attention and count prior embedding for smoke segmentation[J].Pattern Recognition，2022，131：108902.
[27] TAO H.A label-relevance multi-direction interaction network with enhanced deformable convolution for forest smoke recognition[J].Expert Systems with Applications，2024，236：121383.
[28] TIAN H D，LI W Q，OGUNBONA P O，et al.Detection and separation of smoke from single image frames[J].IEEE Transactions on Image Processing，2018，27（3）：1164-1177.
[29] WU X H，LU X B，LEUNG H.Video smoke separation and detection via sparse representation[J].Neurocomputing，2019，360：61-74.
[30] JIA Y，DU H R，WANG H J，et al.Automatic early smoke segmentation based on conditional generative adversarial networks[J].Optik，2019，193：162879.
[31] WANG S D，HE Y P，ZOU J J，et al.Early smoke detection in video using swaying and diffusion feature[J].Journal of Intelligent & Fuzzy Systems：Applications in Engineering and Technology，2014，26（1）：267-275.
[32] CHOLLET F.Xception：deep learning with depthwise separable convolutions[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2017：1800-1807.
[33] SIMONYAN K，ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[C]//Proceedings of the 3rd International Conference on Learning Representations，2015.
[34] HE K M，ZHANG X Y，REN S Q，et al.Deep residual learning for image recognition[C]//Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2016：770-778.
[35] HUANG G，LIU Z，VAN DER MAATEN L，et al.Densely connected convolutional networks[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2017：2261-2269.
[36] YU F，KOLTUN V.Multi-scale context aggregation by dilated convolutions[C]//Proceedings of the 4th International Conference on Learning Representations，2016.
[37] YANG M K，YU K，ZHANG C，et al.DenseASPP for semantic segmentation in street scenes[C]//Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2018：3684-3692.
[38] KIRKLAND E J.Bilinear interpolation[M]//Advanced computing in electron microscopy.Boston：Springer，2010：261-263.
[39] OLIVIER R，CAO H Q.Nearest neighbor value interpolation[J].International Journal of Advanced Computer Science and Applications，2012，3（4）：25-30.
[40] 王潇，李子琦，高涛，等.基于无损跨尺度特征融合的交通目标检测算法[J].中国公路学报，2023，36（9）：315-325.
WANG X，LI Z Q，GAO T，et al.Traffic target detection algorithm based on non-loss cross-scale feature fusion[J].China Journal of Highway and Transport，2023，36（9）：315-325.
[41] LI X Y，SUN X F，MENG Y X，et al.Dice loss for data-imbalanced NLP tasks[EB/OL].[2024-10-15].https：//arxiv.org/abs/1911.02855.
[42] YAN S Y，ZHANG J，BARNES N.Transmission-guided Bayesian generative model for smoke segmentation[C]//Proceedings of the 36th AAAI Conference on Artificial Intelligence.Palo Alto：AAAI，2022：3009-3017.
[43] 李红光，于若男，丁文锐.基于深度学习的小目标检测研究进展[J].航空学报，2021，42（7）：100-118.
LI H G，YU R N，DING W R.Research development of small object traching based on deep learning[J].Acta Aeronautica et Astronautica Sinica，2021，42（7）：100-118.
[44] RONNEBERGER O，FISCHER P，BROX T.U-Net：convolutional networks for biomedical image segmentation[C]//Proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention.Cham：Springer，2015：234-241.
[45] XIE E Z，WANG W H，YU Z D，et al.SegFormer：simple and efficient design for semantic segmentation with transformers[C]//Advances in Neural Information Processing Systems 34，2021：12077-12090.
[46] ZHAO H S，SHI J P，QI X J，et al.Pyramid scene parsing network[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway：IEEE，2017：6230-6239.
[47] WANG J D，SUN K，CHENG T H，et al.Deep high-resolution representation learning for visual recognition[J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2021，43（10）：3349-3364.
[48] YUAN F N，ZHANG L，XIA X，et al.A gated recurrent network with dual classification assistance for smoke semantic segmentation[J].IEEE Transactions on Image Processing，2021，30：4409-4422.
[49] CHEN L C，ZHU Y K，PAPANDREOU G，et al.Encoder-decoder with atrous separable convolution for semantic image segmentation[C]//Proceedings of the 15th European Conference on Computer Vision.Cham：Springer，2018：833-851.
[50] ZHOU F R，WEN G，PAN H，et al.A boundary enhancement and pixel alignment based smoke segmentation network[J].Frontiers in Physics，2023，11：1206944.
[51] YAO L J，ZHAO H T，PENG J C，et al.FoSp：focus and separation network for early smoke segmentation[C]//Proceedings of the 38th AAAI Conference on Artificial Intelligence.Palo Alto：AAAI，2024：6621-6629.