CAJ | 학술논문

提出了一种无人机搭载普通相机的林火识别技术，其是一种低成本无人机林火监测方法。本研究以旋翼无人机为载体，通过在南京森林警察学院院内的两块实验场地（林地、无林土丘）进行点火试验，以机载摄像机拍摄的森林视频图像建立了基于混合高斯背景模型和颜色模型的多级火灾隐患特征验证算法。在同一区域，结合地面调查数据，对无人机搭载普通相机林火识别技术精度进行检验。数据表明，在混合高斯模型得到候选火焰像素的基础上，通过试验设置最优阈值，采用归一化互相关方法设定相似度阈值为0.08，可实现对火焰特征的检测与识别。通过低成本的机载普通相机能较快地识别火灾隐患，降低误检率，可为相关研究和实际应用提供参考。
제출료일충무인궤탑재보통상궤적림화식별기술，기시일충저성본무인궤림화감측방법。본연구이선익무인궤위재체，통과재남경삼림경찰학원원내적량괴실험장지（임지、무림토구）진행점화시험，이궤재섭상궤박섭적삼림시빈도상건립료기우혼합고사배경모형화안색모형적다급화재은환특정험증산법。재동일구역，결합지면조사수거，대무인궤탑재보통상궤림화식별기술정도진행검험。수거표명，재혼합고사모형득도후선화염상소적기출상，통과시험설치최우역치，채용귀일화호상관방법설정상사도역치위0.08，가실현대화염특정적검측여식별。통과저성본적궤재보통상궤능교쾌지식별화재은환，강저오검솔，가위상관연구화실제응용제공삼고。
To avoid the high cost of a infrared and hyperspectral imaging system,a research method for identifying forest fire by using Unmanned Aircraft Vehicle (UAV) equipped with an ordinary camera is proposed,which costs less. Based on the ignition experiments made at two sites (forest land and non-forest land)located in Nanjing Forest Police College,multi-level hidden fire danger feature verification algorithm is built with Gaussian Mixture Model and RGB Color Model based on forest video images shot with an airborne camera. Based on the ground survey data in the same region,the accuracy of forest fire identifying technology using an ordinary camera mounted on an unmanned aircraft vehicle. The data show that he detection and identification of the fire features can be realized through the optimal threshold of 0.08 set by the normalized cross-correlation method on the basis of candidate flame pixel derived from Gaussian Mixture Model,which indicates that a low-cost airborne ordinary camera can rapidly identify hidden fire danger and reduce false detection rate,which can provide reference for relevant research and practical application.