主管：中华人民共和国应急管理部
主办：应急管理部天津消防研究所
ISSN 1009-0029  CN 12-1311/TU
消防科学与技术

消防科学与技术 ›› 2026, Vol. 45 ›› Issue (6): 128-135.DOI: 10.20168/j.1009-0029.2026.06.0128.08

• • 上一篇    下一篇

基于CO2的光谱雷达探测在林火识别中的应用研究

石宽1,2, 游政3, 白夜1,2, 齐方忠1,2, 武英达1,2, 宫大鹏1,2, 袁斌豪3, 尚明宝4   

  1. （1.中国消防救援学院，北京 102202； 2.森林草原火灾风险防控应急管理部重点实验室，北京 102202； 3.江苏大元物联网科技有限公司，江苏 宜兴 214222； 4.四川消防救援机动总队，四川 成都 610043）
  • 收稿日期:2025-05-02 修回日期:2025-07-02 出版日期:2026-06-15 发布日期:2026-06-15
  • 作者简介:石宽，中国消防救援学院森林草原防灭火研究中心，讲师，主要从事森林火灾监测预警研究，北京市昌平区南口镇南雁路4号25号楼，102202。
  • 基金资助:
    森林草原火灾风险防控应急管理部重点实验室开放课题（FGFRP202502）；国家重点研发计划项目（2020YFC1511600）

Research on the application of CO₂-based spectral LiDAR detection in forest fire identification

Shi Kuan1,2, You Zheng3, Bai Ye1,2, Qi Fangzhong1,2, Wu Yingda1,2, Gong Dapeng1,2, Yuan Binhao3, Shang Mingbao4   

  1. (1. China Fire and Rescue Institute, Beijing 102202, China; 2. Emergency Management Department Key Laboratory of Forest Grassland Fire Risk Prevention and Control, Beijing 102202, China; 3. Jiangsu Dayuan IoT Technology Co., Ltd., Yixing Jiangsu 214222, China; 4. Sichuan Fire and Rescue Mobile Brigade, Chengdu Sichuan 610043, China)
  • Received:2025-05-02 Revised:2025-07-02 Online:2026-06-15 Published:2026-06-15

摘要： 传统林火监测技术，如视频监控、红外热成像、卫星遥感与激光雷达等，在复杂地形、多变气候及远距离监测条件下均存在一定局限性，难以实现对初期火情的高精度、广覆盖和全天候探测。本研究选用制冷型HgCdTe红外探测器，聚焦CO₂在4.3 μm波段的强辐射特性，构建了具备360°×80°空间覆盖能力的全景扫描光学结构，并结合步进电机控制的双轴转动系统，实现空间矩阵采集。系统集成高速模数采样平台，由现场可编程门阵列（FPGA）完成144 s警戒区域广域探测视场扫描及4 608万次AD高速采样，并采用矩阵转置与列逆序等算法，将空间信号转化为全景光谱图像。依托远程图像分析平台，基于卷积神经网络对CO2特征区域进行识别与量化评估，实现AI火情精准判别；结合雷达测距与基于北斗导航系统的空间解算算法，实现火情定位。通过在野外搭建小型试验火源，并在不同气象条件下开展实地探测，结果表明，基于CO2的光谱雷达林火探测可在2.05 km距离下，实现对50 cm×80 cm小型火源的精准探测，火源识别总体准确率达到97.31%，定位误差为25.2～345.7 m。基于CO2红外辐射特征的光谱雷达监测技术具有探测灵敏度高、空间分辨率强及智能识别程度高等优点，为林火早期识别提供了新的解决路径。未来可与卫星遥感、视频监测等手段协同，构建天空地一体化、多源信息融合的林火监测体系，为我国森林草原火灾应急响应与生态安全保障提供技术支撑。

关键词: 林火监测, 光谱雷达, 红外探测, CO2辐射, 图像识别, 信息融合

Abstract: Traditional forest fire monitoring technologies, such as video surveillance, infrared thermal imaging, satellite remote sensing, and LiDAR, have certain limitations under conditions of complex terrain, variable weather, and long-range observation, which hinder high-precision, wide-coverage, and all-weather early fire detection. This study employed a cooled HgCdTe infrared detector, targeting the strong infrared radiative characteristics of CO₂ at the 4.3 μm wavelength. A panoramic scanning optical structure with a spatial coverage of 360° × 80° was constructed, and a dual-axis rotation system driven by stepper motors was implemented to enable spatial matrix data acquisition. The system integrated a high-speed analog-to-digital sampling platform, in which a Field-Programmable Gate Array (FPGA) performed a 144 s wide-area scan of the surveillance field and executed 46.08 million high-speed AD sampling cycles. Spatial signals were converted into panoramic spectral images using matrix transposition and column-reverse algorithms. A convolutional neural network, deployed on the cloud-based remote image analysis platform, was used to identify and quantify CO₂-featured regions, achieving AI-based precise fire detection. Fire localization was achieved through radar ranging combined with spatial positioning algorithms based on the BeiDou Navigation Satellite System. Field experiments were conducted using small-scale test fire sources under various weather conditions. The results showed that the CO₂-based spectral LiDAR system could accurately detect a small fire source (50 cm×80 cm) at a distance of 2.05 km. The system achieved an overall identification accuracy of 97.31%, with localization errors ranging from 25.2 m to 345.7 m. Spectral LiDAR monitoring technology based on CO₂ infrared radiation features offers high detection sensitivity, strong spatial resolution, and a high level of intelligent recognition, providing a novel solution for early forest fire identification. In the future, it can be integrated with satellite remote sensing and video surveillance to establish a space-air-ground integrated, multi-source forest fire monitoring system, thereby providing technical support for emergency response and ecological security in China's forest and grassland fire management.

Key words: forest fire monitoring, spectral lidar, infrared detection, CO? radiation, image recognition, information fusion