基于随机森林模型的岸线不透水面提取方法研究

Research on Coastline Impervious Surface Extraction Method Based on Random Forest Model

以南京长江沿岸保护区域为研究对象,利用2016年和2019年两期Sentinel-2影像构造多种遥感指数;再利用随机森林模型进行不透水面信息提取。结果表明,2016年的总体精度为88.25%,Kappa系数为0.76;2019年的总体精度为90.51%,Kappa系数为0.80,说明该方法具有较好的精度,能准确反映不透水面的空间分布。结合提取结果进行时空变化分析发现,2016-2019年南京长江沿线不透水面整体呈减少趋势,总面积减少2.784 km^(2),年均减少0.928 km^(2),其中大部分是由工矿用地转化为绿地以及港口码头消失导致的。长江沿线的不透水面提取和变化检测分析有助于全面、细致和真实地监测河道岸线情况,对于南京市开展长江岸线研究以及开发长江岸线资源和可持续利用具有举足轻重的意义。

We used Sentinel-2 images in 2016 and 2019 to construct multiple remote sensing indices,and applied random forest model to extract impervious surface in the Nanjing Yangtze River protection area.The results show that the overall accuracy and Kappa coefficient in 2016 are 88.25%and 0.76,and the overall accuracy and Kappa coefficient in 2019 are 90.51%and 0.80,indicating that the method has good accuracy and can accurately reflect the spatial distribution of impervious surfaces.Based on the extraction results,we analyzed the impervious surface spatiotemporal changes.The impervious surface along the Yangtze River in Nanjing shows an overall decreasing trend from 2016 to 2019.The total area decreased by 2.784 km^(2),with an average annual decrease of 0.928 km^(2),most of which was caused by the conversion from industrial and mining land to green spaces and the disappearance of ports and docks.The impervious surface extraction and change detection analysis along the Yangtze River is beneficial for the comprehensive and accurate river coastline condition monitoring,and important for the research and sustainable resource utilization along the Nanjing Yangtze River coastline.