基于卷积神经网络的林火小目标和烟雾检测模型

薛震洋; 林海峰; 焦万果

doi:10.12302/j.issn.1000-2006.202308047

薛震洋, 林海峰, 焦万果

南京林业大学学报（自然科学版） ›› 2025, Vol. 49 ›› Issue (1) : 225-234.

PDF(32293 KB)

国家林草科技领军期刊
中国精品科技期刊
中国高校百佳科技期刊
江苏省新闻出版政府奖期刊奖
RCCSE林学权威期刊（A+）
CSCD核心期刊
Scopus数据库收录期刊
中文核心期刊
SCD核心期刊

作者加群：102861116

微信公众号：南京林业大学学报

高级检索

PDF(32293 KB)

南京林业大学学报（自然科学版） ›› 2025, Vol. 49 ›› Issue (1) : 225-234. DOI: 10.12302/j.issn.1000-2006.202308047

研究论文

基于卷积神经网络的林火小目标和烟雾检测模型

作者信息 +

Convolutional neural network-based small target and smoke detection model for forest fires

Author information +

文章历史 +

摘要

【目的】由于火焰和烟雾都是动态目标,使得航拍视角下的森林火焰小目标和烟雾检测变得富有挑战性。基于卷积神经网络的林火小目标和烟雾检测模型构建对于平衡森林火灾小目标和烟雾目标的参数量和检测精度具有重要意义。【方法】根据林火检测实时性要求,对基于高实时性的FasterNet进行改进,并构建了FRSCnet的主干网络层。利用分组洗牌卷积(group shuffle convolution,GSConv)构建了FRSCnet的颈部网络层。在头部网络中,引入了卷积块注意力模块(CBAM)注意力机制。最后,在颈部网络的第1层添加了感受野阻滞(receptive field block,RFB)特征提取模块。【结果】FasterNet采用独特的部分卷积(partial convolution,PConv)技术,有效减少了冗余计算和内存访问,从而保证了模型在处理航拍图像时的高速检测性能。GSConv能够平衡模型的参数量和特征融合性能,从而在保证检测精度的同时减少了模型的复杂度。通过CBAM注意力机制,在尽可能少地增加参数量的情况下,模型能够更好地关注森林火灾小目标和烟雾目标,提高检测性能。RFB特征提取模块,可增大模型的感受野,更好地提取航拍视角下的森林火灾小目标和烟雾目标。实验结果表明,提出的FRSCnet模型的平均精度(mAP@0.5)达到了89.2%,比YOLOv7-tiny和YOLOv5-s模型分别提高了2.9%和5.3%,但是参数量比YOLOv7-tiny和YOLOv5-s模型低。【结论】提出的FRSCnet林火小目标和烟雾目标检测模型,在参数量和检测精度之间取得了较好的平衡。

Abstract

【Objective】Since both flame and smoke are dynamic targets, it makes forest fire small targets and smoke detection in aerial view challenging. The construction of forest fire small target and smoke detection model based on convolutional neural network has positive significance in balancing the number of parameters and detection accuracy of forest fire small target and smoke target.【Method】For the real-time requirements, the high real-time based FasterNet is improved and the backbone network layer of FRSCnet was constructed. The neck network layer of FRSCnet was constructed using group shuffle convolution (GSConv). In the head network, the convolutional block attention module (CBAM) was introduced. Finally, receptive field block (RFB) feature extraction module was added to the first layer of the neck network. 【Result】FasterNet employs a unique partial convolution (PConv) technique that effectively reduce redundant computations and memory accesses, thus ensuring the high-speed detection performance of the model when processing aerial images.GSConv is able to balance the number of parameters of the model with the performance of the feature fusion, thus reducing the complexity of the model while ensuring the detection accuracy. GSConv is able to balance the number of parameters and feature fusion performance of the model, thus reducing the complexity of the model while ensuring detection accuracy. With the CBAM attention mechanism, the model is able to better focus on forest fire small targets and smoke targets to improve the detection performance while increasing the number of parameters as little as possible. To increase the receptive field of the model to better extract forest fire small targets and smoke targets in aerial view.RFB (receptive field block) feature extraction module, which can increase the receptive field of the model to better extract forest fire small targets and smoke targets in aerial view. The experimental results show that the proposed FRSCnet model achieves a performance of 89.2% in terms of average accuracy (mAP@0.5), which is 2.9% and 5.3% better than YOLOv7-tiny and YOLOv5-s, respectively, but the number of parameter counts is lower than YOLOv7-tiny and YOLOv5-s. 【Conclusion】The proposed FRSCnet forest fire small target and smoke target detection model strikes a good balance between the number of parameters and detection accuracy.

导出引用

薛震洋, 林海峰, 焦万果. 基于卷积神经网络的林火小目标和烟雾检测模型[J]. 南京林业大学学报（自然科学版）. 2025, 49(1): 225-234 https://doi.org/10.12302/j.issn.1000-2006.202308047

XUE Zhenyang, LIN Haifeng, JIAO Wanguo. Convolutional neural network-based small target and smoke detection model for forest fires[J]. JOURNAL OF NANJING FORESTRY UNIVERSITY. 2025, 49(1): 225-234 https://doi.org/10.12302/j.issn.1000-2006.202308047

中图分类号： TP391.4

参考文献

列表( 原文顺序 | 文献年度倒序 | 文中引用次数倒序 ) 可视化分析

[1]	WANG Y, YANG X B, ZHANG L, et al. Individual tree segmentation and tree-counting using supervised clustering[J]. Comput Electron Agric, 2023, 205:107629.DOI: 10.1016/j.compag.2023.107629. 本文引用 [1]

[2]	YANG X B, HUA Z C, ZHANG L, et al. Preferred vector machine for forest fire detection[J]. Pattern Recognit, 2023, 143:109722.DOI: 10.1016/j.patcog.2023.109722. 本文引用 [1]

[3]	ZHANG J G, LI W B, HAN N, et al. Forest fire detection system based on a ZigBee wireless sensor network[J]. Front For China, 2008, 3(3):369-374.DOI: 10.1007/s11461-008-0054-3. 本文引用 [1]

[4]	伍小洁, 陈利明, 张洁, 等. 林火监测技术分析与综合应用[J]. 卫星应用, 2017(5):24-28. WU X J, CHEN L M, ZHANG J, et al. Analysis and comprehensive application of forest fire monitoring technology[J]. Satell Appl, 2017(5):24-28.DOI: 10.3969/j.issn.1674-9030.2017.05.005. 本文引用 [1]

[5]	XU Y S, REN F Y, HE T, et al. Real-time routing in wireless sensor networks[J]. ACM Trans Sen Netw, 2013, 9(3):1-24.DOI: 10.1145/2480730.2480738. 本文引用 [1]

[6]	XU R, LIN H, LU, K, et al. A forest fire detection system based on ensemble learning[J]. Forests 2021, 12:217. DOI:10.3390/f12020217. 本文引用 [1]

[7]	MARBACH G, LOEPFE M, BRUPBACHER T. An image processing technique for fire detection in video images[J]. Fire Saf J, 2006, 41(4):285-289.DOI: 10.1016/j.firesaf.2006.02.001. 本文引用 [1]

[8]	杨秋霞, 罗传文. 基于稀疏表示的森林火灾火焰识别研究[J]. 安徽农业科学, 2014, 42(30):10777-10779. YANG Q X, LUO C W. Recognition of forest fire flame based on sparse representation[J]. J Anhui Agric Sci, 2014, 42(30):10777-10779.DOI: 10.13989/j.cnki.0517-6611.2014.30.121. 本文引用 [1]

[9]	MUHAMMAD K, AHMAD J, MEHMOOD I, et al. Convolutional neural networks based fire detection in surveillance videos[J]. IEEE Access, 2018, 6:18174-18183.DOI: 10.1109/access.2018.2812835. 本文引用 [1]

[10]	PAN H Y, BADAWI D, ZHANG X, et al. Additive neural network for forest fire detection[J]. Signal Image Video Process, 2020, 14(4):675-682.DOI: 10.1007/s11760-019-01600-7. 本文引用 [1]

[11]	HU Y W, ZHAN J L, ZHOU G X, et al. Fast forest fire smoke detection using MVMNet[J]. Knowl Based Syst, 2022, 241:108219.DOI: 10.1016/j.knosys.2022.108219. 本文引用 [1]

[12]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition, June 23-28,2014,Columbus,OH,USA.IEEE, 2014:580-587.DOI: 10.1109/CVPR.2014.81. 本文引用 [1]

[13]	REN S, HE K, GIRSHI C K, et al. Faster R-CNN: towards real-time object detection with region proposal networks[C]// Proceedings of the Advances in Neural Information Processing Systems, December 7-12, 2015,Montreal, QC, Canada. DOI: 10.1109/tpami.2016.2577031. 本文引用 [1]

[14]	HE K M, GKIOXARI G, DOLLÁR P, et al. Mask R-CNN[C]// 2017 IEEE International Conference on Computer Vision (ICCV), October 22-29,2017,Venice,Italy.IEEE, 2017:2980-2988.DOI: 10.1109/ICCV.2017.322. 本文引用 [1]

[15]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multiBox detector[C]// European Conference on Computer Vision. Springer: Cham, Switzerland, 2016. DOI: arxiv-1512.02325. 本文引用 [1]

[16]	BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4:optimal speed and accuracy of object detection[EB/OL]. 2020:2004.10934. https://arxiv.org/abs/2004.10934v1. https://arxiv.org/abs/2004.10934v1 本文引用 [1]

[17]	TAN M X, PANG R M, LE Q V. EfficientDet:scalable and efficient object detection[EB/OL]. 2019[2023-03-16]. https://arxiv.org/abs/1911.09070v7. https://arxiv.org/abs/1911.09070v7 本文引用 [1]

[18]	王寅凯, 曹磊, 钱佳晨, 等. 一种改进YOLOv5的多尺度像素林火识别算法[J]. 林业工程学报, 2023, 8(2):159-165. WANG Y K, CAO L, QIAN J C, et al. Multi-scale forest fire recognition using improved YOLOv5 algorithm[J]. J For Eng, 2023, 8(2):159-165.DOI: 10.13360/j.issn.2096-1359.202208019. 本文引用 [1]

[19]	叶铭亮, 周慧英, 李建军. 基于改进Swin Transformer的森林火灾检测算法[J]. 中南林业科技大学学报, 2022, 42(8):101-110. YE M L, ZHOU H Y, LI J J. Forest fire detection algorithm based on an improved Swin Transformer[J]. J Cent South Univ For Technol, 2022, 42(8):101-110.DOI: 10.14067/j.cnki.1673-923x.2022.08.010. 本文引用 [1]

[20]	蔡魏斌. 基于卷积神经网络的视频烟雾监测系统研究[D]. 西安: 长安大学, 2022. CAI W B. Research on video smoke monitoring system based on convolutional neural network[D]. Xi’an: Chang’an University, 2022.DOI: 10.26976/d.cnki.gchau.2022.001852. 本文引用 [1]

[21]	张倩, 周平平, 王公堂, 等. 基于合成图像的Faster R-CNN森林火灾烟雾检测[J]. 山东师范大学学报(自然科学版), 2019, 34(2):180-185. ZHANG Q, ZHOU P P, WANG G T, et al. Faster R-CNN forest fire smoke detection based on synthetic images[J]. J Shandong Norm Univ (Nat Sci), 2019, 34(2):180-185.DOI: 10.3969/j.issn.1001-4748.2019.02.010. 本文引用 [1]

[22]	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 (CVPR), June 17-24,2023,Vancouver,BC,Canada.IEEE, 2023:12021-12031.DOI: 10.1109/CVPR52729.2023.01157. 本文引用 [1]

[23]	LI H, LI J, WEI H, et al. Slim-neck by GSConv: a better design paradigm of detector architectures for autonomous vehicles[J]. arXiv preprint arXiv, 2022, 2206:02424.DOI: arxiv-2206.02424. 本文引用 [1]

[24]	LIU S T, HUANG D, WANG Y H. Receptive field block net for accurate and fast object detection[M]// Lecture notes in computer science. Cham: Springer International Publishing, 2018:404-419.DOI: 10.1007/978-3-030-01252-6_24. 本文引用 [1]

[25]	WOO S, PARK J, LEE J Y, et al. CBAM:convolutional block attention module[M]//Lecture notes in computer science. Cham: Springer International Publishing, 2018:3-19.DOI: 10.1007/978-3-030-01234-2_1. 本文引用 [1]

[26]	MEHTA S, RASTEGARI M. MobileViT:light-weight,general-purpose,and mobile-friendly vision transformer[EB/OL]. 2021[2023-04-15]. https://arxiv.org/abs/2110.02178v2. https://arxiv.org/abs/2110.02178v2 本文引用 [1]

[27]	WANG C Y, MARK LIAO H Y, WU Y H, et al. CSPNet:a new backbone that can enhance learning capability of CNN[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), June 14-19,2020,Seattle,WA,USA.IEEE, 2020:1571-1580.DOI: 10.1109/CVPRW50498.2020.00203. 本文引用 [1]