基于多光谱遥感图像的火点检测技术  

Fire Spot Detection Technology Based on Multispectral Remote Sensing Images

作　　者：冯希胤 刘梦 姜艳 李路 范雪松 FENG Xiyin;LIU Meng;JIANG Yan;LI Lu;FAN Xuesong(Wuchang Shipbuilding Industry Group Co.,Ltd.,Wuhan 430416,China)

机构地区：[1]武昌船舶重工集团有限公司,湖北武汉430416

出　　处：《无线电工程》2021年第11期1195-1201,共7页Radio Engineering

基　　金：2019工业互联网创新发展工程项目(AK190015K)。

摘　　要：为探究遥感卫星在监测火灾方面的应用,通过对Landsat 8图像和高分四号卫星图像波段进行分析,采用“劈窗法”构建高分四号(GF-4)卫星中波红外波段数据的着火点检测算法。对Landsat 8和GF-4卫星的图像数据先进行预处理,对每个像元进行检测,根据该像元的特定波段的反射率和亮温值是否满足所设阈值决定是否将其判断为所需像元。对火点检测结果进行分析,使用中波红外波段的亮点和有色光波段的烟雾共同判断火点检测的准确性。验证结果表明,出火点检测准确率在90%以上。In order to explore the application of remote sensing satellites in monitoring fires,the Landsat 8 image and the Gaofen-4 satellite image band is analyzed.The“split window method”is used to construct a fire point detection algorithm for the mid-wave infrared band data of the Gaofen-4 satellite.The image data of Landsat 8 and Gaofen-4 satellites are preprocessed first,and then each pixel is detected.According to whether the reflectivity and brightness temperature of the specific band of the pixel meet the set threshold,it is determined whether it is judged as the desired pixel.The results of the fire spot detection are analyzed,and the bright spots in the mid-wave infrared band and the smoke in the colored light band are used to jointly judge the accuracy of the fire spot detection.The experimental results show that the calculated accuracy of fire spot detection is above 90%.

关 键 词：高分四号 Landsat 8 阈值分割 火点检测 多光谱 遥感图像 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]