城市不透水面遥感信息提取精度对比研究
Comparison of the extraction accuracy of urban impervious surfaces information based on remote sensing
王厚望1
作者信息
- 1. 上海市测绘院,上海 200063;自然资源部超大城市自然资源时空大数据分析应用重点实验室,上海 200063
- 折叠
摘要
不透水面是反映城市工程建设水平和评价城市生态环境质量的有效指标,因此,高效提取不透水面对城市扩张监测和生态系统建设有重要意义.本文以上海市城区为研究对象,首先利用D-InSAR技术对Sentinel-1A影像进行差分干涉处理,获取相干系数图;然后,与Landsat 8影像进行波段组合;最后,利用多种机器学习算法提取不透水面,并进行精度对比.结果表明,相较于单一Landsat 8影像,加入相干系数图后,最大似然(ML)算法、支持向量机(SVM)算法、分类与回归树(CART)算法和随机森林(RF)算法提取不透水面的精度均有明显提高.其中,RF算法表现最好,总体精度达到89.33％.
Abstract
Impervious surface is an important indicator to reflect the level of urban engineering construc-tion and it can be used to evaluate the quality of urban ecological environment.Therefore,efficient extraction of impervious surface is of great significance for urban expansion monitoring and ecosystem construction.D-InSAR technology is used in this paper to perform differential interferometry on Sentinel-1A images,generate coherence coefficient map,and combine bands with Landsat 8 images by taking the urban area of Shanghai City as the study area.Then a variety of machine learning algorithms are used to extract impervious surface information and the accuracy is compared.The results show that compared with a single Landsat 8 image,the maximum likelihood algorithm,SVM algorithm,CART algorithm and RF algorithm can be used to greatly improve the extraction accuracy of impervious surface information,and among which the RF algorithm can get the best accuracy,its overall accuracy can be 89.33％if the coherence coefficient map is added.
关键词
不透水面/D-InSAR/相干系数图/机器学习算法Key words
impervious surface/D-InSAR/coherence coefficient map/machine learning algorithm引用本文复制引用
基金项目
上海市2021年度"科技创新行动计划"社会发展科技攻关项目(21DZ1204100)
出版年
2024
NETL
NSTL
维普
万方数据