基于双光谱的森林烟火预警监测研究
Study on early warning and monitoring of forest fireworks based on bispectrum
[目的]森林火灾危害着自然资源、经济发展和人类安全。烟雾是森林火灾的早期显著特征,能够作为林火的有效监测对象。本研究旨在推进森林烟火的智能识别,构建完备的林火综合管理平台,达成迅速、高效的预警监测。[方法]本研究采纳森林火灾早期阶段的识别特征,多角度对森林烟火业务的痛点需求进行分析,对林火监控设备进行成像能力测试、监护范围及定位能力测试、防护性能检验及测试、森林烟火智能识别能力测试。其中,对森林防火专业 AI 摄像机和高精度球台等林火视频感知设备进行烟火智能识别能力测试,分别测量不同点烟距离、点火距离的识别能力以及离线火情识别能力;通过“红外坐标映射+可见光辅助判别”的模式,结合多光谱算法,组建复合部署结构,建设森林防火综合管理平台。[结果]本研究累积百万级烟火图像样本和非烟火样本数据,构建可见光火情样本库和热成像数据样本库;通过采用先进的前端烟火识别算法和边端过滤算法,使林火烟雾智能识别漏报率小于千分之一,误报率达到每万公顷每日小于2次;森林防火综合管理平台以智能识别、多端联合等特点,能够有效达成森林火灾的监测预警,并指导消防队伍及时救援火情。[结论]双光谱技术指导下,森林烟火探测装备的合理化、高效化应用落实于预警监测实践中,能够减少森林火灾的漏报率和误报率,满足森林烟火业务的痛点需求。
[Objective]Forest fires are endangering natural resources,economic development and human security.Smoke is an early significant feature of forest fire,which can be used as an effective monitoring object of forest fire.The purpose of this study is to promote the intelligent identification of forest fireworks,build a comprehensive forest fire management platform,and achieve rapid and efficient early warning and monitoring.[Method]In this study,the identification characteristics in the early stage of forest fire were adopted,and the pain demand of forest fireworks business was analyzed from multiple angles.The imaging ability test,monitoring range and positioning ability test,protective performance test and test,and intelligent identification ability test of forest fireworks were carried out for forest fire monitoring equipment.Among them,the intelligent recognition ability of forest fire video sensing equipment such as forest fire prevention professional AI camera and high-precision table is tested,and the recognition ability of different cigarette lighting distances and ignition distances and the off-line fire recognition ability are measured respectively;Through the mode of"infrared coordinate mapping+visible light aided discrimination"and multi-spectral algorithm,a composite deployment structure is established to build a comprehensive forest fire prevention management platform.[Result]In this study,millions of pyrotechnic image samples and non-pyrotechnic sample data were accumulated,and the visible light fire sample database and thermal imaging data sample database were constructed.By adopting advanced front-end smoke identification algorithm and edge-end filtering algorithm,the missing rate of intelligent identification of forest fire smoke is less than one thousandth,and the false alarm rate is less than two times per ten thousand hectares per day.With the characteristics of intelligent identification and multi-terminal combination,the integrated management platform of forest fire prevention can effectively achieve the monitoring and early warning of forest fires and guide the fire brigade to rescue the fire in time.[Conclusion]Under the guidance of dual-spectrum technology,the rationalization and efficient application of forest fireworks detection equipment in early warning and monitoring practice can reduce the rate of missing report and false report of forest fires and meet the pain point demand of forest fireworks business.
陈卓;侯冬青;刘海洋;梁保柱;党祯;郑明章;
滇西科技师范学院智能与信息工程学院 临沧 677000##清华大学公共安全研究院安全科学学院 北京 100084;滇西科技师范学院智能与信息工程学院 临沧 677000;滇西科技师范学院智能与信息工程学院 临沧 677000##清华大学公共安全研究院安全科学学院 北京 100084;滇西科技师范学院智能与信息工程学院 临沧 677000;中国移动通信集团云南有限公司大理分公司 大理 671001;吉林省林业勘察设计研究院 长春 130022;
S762;TP3
Smart forestry Forest fire Smoke recognition Bispectrum Early warning monitoring
1367-1379 / 13
评论