摘要
利用遥感技术获取红树林的空间分布信息,可弥补常规调查的不足,对保护利用红树林具有重要意义。本文利用Google Earth Engine(GEE)云平台获取Sentinel-2卫星影像作为数据源,提取遥感光谱特征及植被指数信息;结合广西沿海地区红树林实地调查数据,采用面向对象、基于像元的随机森林方法识别、提取研究区红树林的空间分布,通过混淆矩阵对比验证两种方法的分类效果。结果表明:(1)通过Sentinel-2提取的红边、绿、近红外及短波红外波段组合而成的植被指数对于红树林的特征识别更为有效;(2)面向对象法和随机森林法提取红树林的生产者精度均低于用户精度;(3)随机森林法各项精度验证指标均高于面向对象法,随机森林分类的用户精度与生产者精度分别为91.8%和81.2%,相比于面向对象,其用户精度与生产者精度分别提高了11.3%和24.7%。基于像元的随机森林方法能提供更准确的红树林分布信息,可为红树林信息提取提供参考依据。
The use of remote sensing technology to obtain the spatial distribution of mangrove forests can make up for the shortcomings of traditional investigation,which was of great significance to the protection and utilization of mangrove forests.Google Earth Engine(GEE)cloud platform was used to obtain Sentinel-2 satellite images as data sources,while the remote sensing spectral characteristics and vegetation index information were extracted,combined with the field survey data in Guangxi coastal area.Two classification methods,object-oriented and pixel-based random forest,were used to identify and extract the spatial distribution of mangrove in the study area.Finally,confusion matrix was used to compare and verify the classification effect of the two methods.The results showed that:(1)The vegetation indices combined with red edge,green,near-infrared and short-wave infrared bands extracted from Sentinel-2 were more effective for the characteristic recognition of mangrove forests.(2)The producer accuracy by object-oriented and random forest method to extract mangrove forests was lower than that of user accuracy.(3)The accuracy of random forest method was higher than that of object-oriented method.The user accuracy and producer accuracy of random forest classification was 91.8%and 81.2%,respectively.The user accuracy and producer accuracy was improved by 11.3%and 24.7%compared with object-oriented,respectively.Pixel-based random forest method could provide more accurate mangrove distribution information,which could provide basis and reference for mangrove information extraction.
作者
袁胜
YUAN Sheng(Guangxi Forest Resources and Environment Monitoring Center,Nanning 530028,Guangxi,China)
出处
《湖南林业科技》
2021年第5期34-39,共6页
Hunan Forestry Science & Technology
