基于极化特征变化的面向对象燃烧区提取  

Object-oriented Combustion Area Extraction Based on Polarization Characteristics Change

作　　者：王晶玉 WANG Jingyu(Heilongjiang Institute of Geomatics Engineering,Harbin 150081,China)

机构地区：[1]黑龙江地理信息工程院,黑龙江哈尔滨150081

出　　处：《测绘与空间地理信息》2023年第S01期195-198,共4页Geomatics & Spatial Information Technology

摘　　要：针对利用SAR影像提取林火燃烧区主要侧重于单极化后向散射系数,而从火灾导致散射机制变化的角度开展双极化燃烧区提取的研究较少,本文提出了一种基于极化特征变化的面向对象燃烧区提取方法。以两景Sentinel-1A影像为数据源,构建极化特征变化参数集,使用多尺度分割方法结合随机森林算法提取燃烧区,并采用混淆矩阵精度评价方法对燃烧区提取结果进行评价。实验结果表明,分类总体精度为89.13%,Kappa系数为0.76,说明极化特征变化参数在燃烧区提取方面具有良好的应用能力。The extraction of forest fire combustion area by SAR image mainly focuses on the single polarization backscattering coeffi-cient,and there are few researches on the extraction of dual polarization combustion area from the perspective of the change of scatter-ing mechanism caused by fire.In this paper,an object-oriented method for extraction of combustion area based on the change of po-larization characteristics is proposed.Two Sentinel-1A images were used as data sources to construct the polarization characteristic change parameter set,and multi-scale segmentation method combined with random forest algorithm was used to extract the combustion area,and confusion matrix accuracy evaluation method was used to evaluate the extraction results of the combustion area.The experi-mental results show that the overall classification accuracy is 89.13%and Kappa coefficient is 0.76,indicating that the polarization characteristic variation parameters have good application ability in the extraction of combustion area.

关 键 词：Sentinel-1A 面向对象 极化特征 林火燃烧区 随机森林 

分 类 号：P237[天文地球—摄影测量与遥感]