基于Himawari-8亚像元分解的林火检测
Detection of Subpixel-level Forest Fire Using Himawari-8 Images Based on Linear Spectral Unmixing Method
郝艳娜 1丁海勇 1张兴强2
作者信息
- 1. 南京信息工程大学遥感与测绘工程学院,南京 210044
- 2. 山东省泰山市气象局,山东泰安 271001
- 折叠
摘要
受空间分辨率的影响,卫星遥感影像普遍存在混合像元,传统的林火监测方法是以像元为单位,容易漏判小火点.针对该问题,文章提出了基于亚像元分解的森林火点检测算法,基于改进的 自适应阈值法进行火点判识,判别影像中林火范围.通过光谱解混确定含典型线状地物的混合像元,利用丰度图像提供的信息将地物类别标签分配到每个亚像元中.基于空间相关性理论构建线性优化模型,获得亚像元定位结果,准确定位到火点所在经纬度.以葵花八号卫星数据为数据源,对本文所提出的亚像元火点检测算法进行了实验.结果表明,基于完全约束最小二乘模型进行林火混合像元分解,可获得火点亚像元级的空间分布细节信息,精度较高,对火点的位置信息提取效果良好.
Abstract
Due to the influence of spatial resolution,mixed pixels are commonly present in satellite remote sensing images.Traditional forest fire monitoring methods are based on pixels,which can easily miss small fire points.To solve this problem,this paper proposes a forest fire point detection algorithm based on subpixel decomposition.Based on an improved adaptive threshold method for fire point identification,identify the forest fire range in the image.The mixed pixels containing typical linear ground objects are determined by spectral unmixing,and the ground objects category labels are assigned to each subpixel using the information provided by the abundant images.A linear optimization model is built based on the spatial correlation theory to obtain the location results of subpixels,and the longitude and latitude of the fire point are accurately located.The sub pixel fire point detection algorithm proposed in this paper is tested using data from the Sunflower 8 satellite as the data source.The results show that the forest fire mixed pixel decomposition based on the fully constrained least squares model can obtain the subpixel level spatial distribution details of the fire point,with high accuracy and good effect on the location information extraction of the fire point.
关键词
Himawari-8静止气象卫星/完全约束最小二乘法/均方根误差/丰度反演/亚像元定位Key words
Himawari-8 stationary meteorological satellite/completely constrained least square method/root mean square error/abundance inversion/subpixel localization引用本文复制引用
基金项目
高分辨率对地观测系统重大专项(30-Y60B01-9003-22/23)
出版年
2023
NETL
NSTL
万方数据