基于轻量级模型的森林火灾图像协同识别方法

刘润滋; 柳辰伟; 吴伟华; 夏文超; 王燕妮; 张琰

doi:null

您当前的位置：

首页 >

文章列表页 >

基于轻量级模型的森林火灾图像协同识别方法

更新时间：2026-03-03

- 基于轻量级模型的森林火灾图像协同识别方法
- Collaborative Forest Fire Identification: A Lightweight Model-Based Approach
- 物联网学报 2026年
- 作者机构：
  
  1.西安建筑科技大学信息与控制工程学院，陕西西安 710399
  2.陕西师范大学物理学与信息技术学院，陕西西安 710119
  3.南京邮电大学通信与信息工程学院，江苏南京 210003
  4.西安电子科技大学通信工程学院，陕西西安 710071
- 作者简介：
  
  吴伟华，whwu@snnu.edu.cn
- 基金信息：
  
  国家自然科学基金项目(61701365);中国科协青年人才托举工程项目(2022QNRC001);陕西省重点研发计划项目(2025CY-YBXM-063);陕西省自然科学基础研究计划项目(2025JC-YBMS-791)
- DOI：
  中图分类号： TP391
- 收稿：2025-07-14，
  
  修回：2025-08-12，
  
  录用：2025-09-01，
- 稿件说明：
移动端阅览
刘润滋, 柳辰伟, 吴伟华, 等. 基于轻量级模型的森林火灾图像协同识别方法[J/OL]. 物联网学报, 2026.

LIU Runzi, LIU Chenwei, WU Weihua, et al. Collaborative Forest Fire Identification: A Lightweight Model-Based Approach[J/OL]. Chinese Journal on Internet of Things, 2026.
刘润滋, 柳辰伟, 吴伟华, 等. 基于轻量级模型的森林火灾图像协同识别方法[J/OL]. 物联网学报, 2026. DOI：

LIU Runzi, LIU Chenwei, WU Weihua, et al. Collaborative Forest Fire Identification: A Lightweight Model-Based Approach[J/OL]. Chinese Journal on Internet of Things, 2026. DOI：

摘要

基于物联网的环境监测体系是森林火灾早期预警的重要基础设施，对遏制灾害蔓延具有决定性作用。然而，部署于野外的监测终端存在通信带宽受限、计算能力弱、能源供给有限的问题，使其既难以将大量高清图像实时回传，又无法在设备端部署并运行复杂的神经网络模型。为此，本文设计了一种基于“端-云”协同的森林火灾识别方法。通过在监测终端上部署轻量级识别模型和卸载模型，终端采集到的大部分图像可以在边缘侧完成本地识别；仅当轻量级模型难以准确识别的少量图像，才通过物联网网络卸载到森林防火指挥中心进行云端识别，从而在满足物联网应用对低延迟、低带宽和低能耗核心需求的同时，兼顾了识别准确性。所设计的轻量级森林火灾图像识别模型结合了Ghost Module和ShuffleNetv2，并引入ECA（Efficient Channel Attention）注意力模块，显著降低了对计算和存储资源的需求。为了提高卸载决策在动态物联网环境下的有效性，提出了一种融合噪声网络的双缓冲区近端策略优化算法（Noisy Double Proximal Policy Optimization，NDPPO）来训练卸载模型。通过对比实验验证了所提出的面向物联网的轻量级森林火灾协同识别方法的有效性。

Abstract

IoT-based environmental monitoring systems are crucial infrastructure for early forest fire warnings

playing a decisive role in preventing disaster spread. However

field-deployed terminals face limitations in communication bandwidth

computational power

and energy supply

hindering both real-time transmission of high-resolution images and on-device deployment of complex neural networks. This paper proposes an edge-cloud collaborative forest fire recognition method. Lightweight recognition and offloading models are deployed on terminals. Most images are processed locally at the edge; only images challenging for the lightweight model are offloaded via the IoT network to a cloud center for recognition. This balances recognition accuracy with the core IoT requirements of low latency

low bandwidth

and energy efficiency. The designed lightweight fire recognition model integrates Ghost Module and ShuffleNetv2

enhanced with an Efficient Channel Attention (ECA) module

significantly reducing computational and storage demands. To improve offloading decision effectiveness in dynamic IoT environments

a Noisy Double Proximal Policy Optimization (NDPPO) algorithm is proposed to train the offloading model. Comparative experiments validate the effectiveness of this lightweight collaborative recognition approach for IoT-based forest fire monitoring.

关键词

Keywords

references

Zhao J , Yue C , Wang J , et al . Forest fire size amplifies postfire land surface warming [J ] . Nature , 2024 , 633 ( 8031 ): 828 - 834 .

凉山州西昌市“ 3 · 30”森林火灾事件调查结果公布 [J ] . 中国消防，2021（ 1 ）： 66 - 67 .

Investigation results of the " 3 · 30 " fire incident in Xichang City, Liangshan Prefecture [J ] . China Fire Rescue, 2021 ( 1 ): 66 – 67 .

Chuvieco E , Aguado I , Salas J , et al . Satellite remote sensing contributions to wildland fire science and management [J ] . Current Forestry Reports , 2020 , 6 : 81 - 96 .

Bonilla-Ormachea K , Cuizaga H , Salcedo E , et al . ForestProtector: An IoT Architecture Integrating Machine Vision and Deep Reinforcement Learning for Efficient Wildfire Monitoring [J ] . arXiv preprint arXiv: 2501.09926 , 2025 .

Chen Y , Zhang Y , Xin J , et al . UAV image-based forest fire detection approach using convolutional neural network [C ] // 2019 14th IEEE conference on industrial electronics and applications (ICIEA) . IEEE , 2019 : 2118 - 2123 .

Zhang A , Zhang A S . Real-time wildfire detection and alerting with a novel machine learning approach [J ] . International Journal of Advanced Computer Science and Applications , 2022 , 13 ( 8 ).

Zheng J , Cao X , Zhang B , et al . Bi-heterogeneous Convolutional Neural Network for UAV-based dynamic scene classification [C ] // 2017 Integrated Communications, Navigation and Surveillance Conference (ICNS) . IEEE , 2017 : 5B4-1-5B4 - 12 .

Allauddin M S , Kiran G S , Kiran G S S R , et al . Development of a surveillance system for forest fire detection and monitoring using drones [C ] // IGARSS 2019-2019 IEEE International Geoscience and Remote Sensing Symposium . IEEE , 2019 : 9361 - 9363 .

Jiao Z , Zhang Y , Mu L , et al . A yolov3-based learning strategy for real-time uav-based forest fire detection [C ] // 2020 Chinese control and decision conference (CCDC) . IEEE , 2020 : 4963 - 4967 .

Yang H , Wang J , Wang J . Efficient Detection of Forest Fire Smoke in UAV Aerial Imagery Based on an Improved Yolov5 Model and Transfer Learning [J ] . Remote Sensing , 2023 , 15 ( 23 ): 5527 .

袁海英 , 成君鹏 , 曾智勇 , 等 . Mobile_BLNet: 基于 Big-Little Net 的轻量级卷积神经网络优化设计 [J ] . 电子学报 , 2023 , 51 ( 1 ): 180 - 191 .

YUAN H Y , CHENG J P , ZENG Z Y , et al . Mobile_BLNet: Optimized design of lightweight convolutional neural network based on Big-Little Net [J ] . Acta Electronica Sinica , 2023 , 51 ( 1 ): 180 - 191 .

Zhou M , Wu L , Liu S , et al . UAV forest fire detection based on lightweight YOLOv5 model [J ] . Multimedia Tools and Applications , 2024 , 83 ( 22 ): 61777 - 61788 .

Xu L , Yang Q , Qin M , et al . Collaborative human recognition with lightweight models in drone-based search and rescue operations [J ] . IEEE Transactions on Vehicular Technology , 2023 , 73 ( 2 ): 1765 - 1776 .

Kang X , Song B , Guo J , et al . Task-oriented image transmission for scene classification in unmanned aerial systems [J ] . IEEE Transactions on Communications , 2022 , 70 ( 8 ): 5181 - 5192 .

Zhou C , Wu W , He H , et al . Deep reinforcement learning for delay-oriented IoT task scheduling in SAGIN [J ] . IEEE Transactions on Wireless Communications , 2020 , 20 ( 2 ): 911 - 925 .

Zhou C , Wu W , He H , et al . Delay-aware IoT task scheduling in space-air-ground integrated network [C ] // 2019 IEEE Global Communications Conference (GLOBECOM) . IEEE , 2019 : 1 - 6 .

Yan K , Shan H , Sun T , et al . Reinforcement learning-based mobile edge computing and transmission scheduling for video surveillance [J ] . IEEE Transactions on Emerging Topics in Computing , 2021 , 10 ( 2 ): 1142 - 1156 .

Han K , Wang Y , Tian Q , et al . Ghostnet: More features from cheap operations [C ] // Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2020 : 1580 - 1589 .

Ma N , Zhang X , Zheng H T , et al . Shufflenetv2: Practical guidelines for efficient cnn architecture design [C ] // Proceedings of the European conference on computer vision (ECCV) . 2018 : 116 - 131 .

丁瑞金 , 高飞飞 , 邢玲 . 基于深度强化学习的物联网智能路由策略 [J ] . 物联网学报 , 2019 , 3 ( 2 ): 56 - 63 .

DING R J , GAO F F , XING L . Intelligent routing strategy in the Internet of things based on deep reinforcement learning . [J ] . Chinese journal on internet of things , 2019 , 3 ( 2 ): 56 - 63 .

Schulman J , Wolski F , Dhariwal P , et al . Proximal policy optimization algorithms [J ] . arXiv preprint arXiv: 1707.06347 , 2017 .

Fortunato M , Azar M G , Piot B , et al . Noisy networks for exploration [J ] . arXiv preprint arXiv: 1706.10295 , 2017 .

Sandler M , Howard A , Zhu M , et al . Mobilenetv2: Invesrted residuals and linear bottlenecks [C ] // Proceedings of the IEEE conference on computer vision and pattern recognition . 2018 : 4510 - 4520 .

Howard A , Sandler M , Chu G , et al . Searching for mobilenetv3 [C ] // Proceedings of the IEEE/CVF international conference on computer vision . 2019 : 1314 - 1324 .

Tan M , Chen B , Pang R , et al . Mnasnet: Platform-aware neural architecture search for mobile [C ] // Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 2019 : 2820 - 2828 .

Saito T , Rehmsmeier M . The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets [J ] . PloS one , 2015 , 10 ( 3 ): e0118432 .

Ling C X , Huang J , Zhang H . AUC: a better measure than accuracy in comparing learning algorithms [C ] // Conference of the canadian society for computational studies of intelligence . Berlin, Heidelberg : Springer Berlin Heidelberg , 2003 : 329 - 341 .

Van Hasselt H , Guez A , Silver D . Deep reinforcement learning with

double q-learning [C ] // Proceedings of the AAAI conference on artificial intelligence . 2016 , 30(1) .

Mnih V , Badia A P , Mirza M , et al . Asynchronous methods for deep reinforcement learning [C ] // International conference on machine learning . PmLR , 2016 : 1928 - 1937 .

Yu C , Velu A , Vinitsky E , et al . The surprising effectiveness of ppo in cooperative multi-agent games [J ] . Advances in neural information processing systems , 2022 , 35 : 24611 - 24624 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于智能手机CSI的轻量级指纹定位模型

共沉淀法制备镁基六铝酸镧粉体的性能研究

低温回火对超大形变冷拔珠光体钢丝显微组织和力学性能的影响

定向凝固糊状区枝晶粗化和二次臂迁移的实验和模拟

热轧板退火工艺对19Cr2Mo1W铁素体不锈钢织构与成形性能的影响