集成夜间灯光数据与Landsat TM影像的不透水面自动提取方法研究

发布时间：2018-04-27 10:23

  本文选题：不透水面 + 自动提取　； 参考：《地球信息科学学报》2017年10期

【摘要】：利用多源遥感数据提取不透水面信息是一个重要的研究方向。针对以往研究中多需要人工选取不透水面样本进行模型训练的问题,本文通过整合夜间灯光遥感与Landsat TM影像中的空间和光谱信息实现了不透水面覆盖范围(Impervious Surface Area,ISA)的自动提取。首先根据夜间灯光的分布来定位ISA聚集的城市区域的位置,分别在城市区域内部和外部自动提取可靠性高的ISA及非ISA样本,然后通过迭代分类提取城市区域的ISA,再以此为样本对城市区域外部进行分类,最后将分类结果整合完成整幅影像的ISA提取流程。应用本方法对美国雪城地区的DMSP/OLS夜间灯光影像上提取了84个城市区域,提取精度大于95%。从中分别选择高ISA密度和低ISA密度的2个城市区域作为ISA提取的测试区,本文方法在城市区域内的ISA提取总体精度与kappa系数分别为88.23%和0.63;在城市区域外部为78.6%和0.54,均优于人工样本选取方法的提取精度,表明该方法能够实现精度稳定且高效的ISA自动提取。
[Abstract]:Extracting impermeable surface information from multi-source remote sensing data is an important research direction. In order to solve the problem of artificial selection of impermeable surface samples for model training in previous studies, this paper realizes automatic extraction of impervious Surface area by integrating spatial and spectral information from night light remote sensing and Landsat TM images. Firstly, according to the distribution of night lights, the location of the urban area where ISA is gathered is located, and the highly reliable ISA and non- samples are automatically extracted from the inner and outer parts of the urban area, respectively. Then the ISAs of the urban area are extracted by iterative classification, and then the ISA extraction process of the whole image is completed by using the ISAs as samples to classify the exterior of the urban area. This method is used to extract 84 urban areas from the DMSP/OLS night light image of the American Snow City area, and the extraction accuracy is more than 95%. Two urban areas with high ISA density and low ISA density were selected as test areas for ISA extraction. The total precision and kappa coefficient of ISA extraction in urban areas were 88.23% and 0.63 respectively, and those outside urban areas were 78.6% and 0.54 respectively, which were better than those of artificial sample selection method. It shows that this method can achieve stable precision and efficient automatic ISA extraction.
【作者单位】： 成都理工大学地球物理学院;浙江工业大学计算机与技术学院;中国科学院遥感与数字地球研究所;中国科学院大学;
【基金】：国家重点研发项目(2017YFB0504204、2016YFB0502502) 国家自然科学基金项目(41301488)
【分类号】：P237
，

本文编号：1810339