多源参考数据下的边坡DEM提取与精度分析

Slope DEM extraction and accuracy analysis based on multi-source reference data

参考DEM是InSAR DEM提取精度分析中不可或缺的重要数据。考虑到当前可供使用的DEM数据层出不穷,如何选择可靠的参考DEM以解决不同地形DEM提取精度的需求,是DEM提取中一个亟待解决的问题。本文分别选择12.5 m的ALOS DEM、30 m的STRM DEM、90 m的STRM DEM作为参考DEM,依据2020年两景Sentinel-1A影像对德兴市部分地区边坡进行DEM数据提取,利用InSAR技术提取实验区的S1ADEM数据,并以ASTGTM DEM数据为参考,使用检测点法随机选择约300个检测点,选择相应的评价指标,分别对不同参考DEM下提取得到的S1ADEM进行精度检验。研究结果显示,使用12.5 m分辨率的ALOS DEM作为外部DEM,提取得到的S1ADEM精度更高,其在实验区A的RMSE为9.412 m,使用90 m的STRM DEM,提取得到的S1ADEM精度更低,其在实验区A的RMSE为10.293m。该实验为不同地形提取DEM提供了指导,能够帮助学者们在进行InSAR实验时更好地选择参考DEM,研究成果可为边坡灾害及预警提供重要的技术支持。

Reference DEM is indispensable and important data in the accuracy analysis of InSAR DEM extraction.Considering the abundance of available DEM data,how to choose reliable reference DEMs to meet the accuracy requirements of DEM extraction for different terrains is an urgent problem in DEM extraction.In this article,ALOS DEM of 12.5 m,STRM DEM of 30 m,and STRM DEM of 90 m were selected as reference DEMs.Based on two Sentinel-1A images from 2020,DEM data on slopes in some areas of Dexing City was extracted.InSAR technology was used to extract S1A DEM data from the experimental area,and ASTGTM DEM data as a reference.About 300 detection points were randomly selected using the detection point method,the corresponding evaluation indicators were selected,and accuracy checks on S1A DEMs extracted from different reference DEMs were performed.The research results show that the extracted S1A DEM has a higher accuracy when ALOS DEM with a resolution of 12.5 m is used as the external DEM,with its RMSE in experimental area A of 9.412 m.The extracted S1A DEM has a relatively lower accuracy when ALOS DEM with a resolution of 90 m is used as the external DEM,with its RMSE in experimental area A of 10.293 m.This experiment guides extracting DEM from different terrains and can help scholars better choose reference DEM when conducting InSAR experiments.The research results can provide important technical support for slope disasters and early warning.