KOMPSAT-3 Digital Elevation Model Correction Based on Point-to-Surface Matching

Hyoseong Lee,Michael Hahn

doi:10.3390/rs11202340

Abstract

In order to generate digital elevation models (DEMs) from high-resolution satellite images, the vendor-provided rational polynomial coefficients (RPCs) are commonly used. This results in a level of accuracy that can be improved by using ground control points (GCPs). The integration of the GCPs into the processing chain is associated with additional effort, since it requires the provision of GCPs as well as the measurement of its image coordinates. In this paper, the authors avoid the measurement of GCP image coordinates and propose a point-to-surface matching method to correct the DEM produced from KOMPSAT-3 satellite images and the provided RPCs. For point-to-surface matching, an existing network of GCPs was used in South Korea, the so-called united control points and the triangulation control points. Practical testing was summarized with the proposed method in which the root mean square error with respect to the horizontal position and the height reduced from 20 m and 6 m to 3 m and 2 m, respectively. This demonstrates that neither image coordinate measurements nor additional GCP point acquisition, e.g., by GPS, are necessary to convert a DEM generated from KOMPSAT-3 images and vendor-provided RPCs into a highly accurate DEM by using existing GCPs and point-to-surface matching.

Highlights

The generation of a high-quality digital elevation model (DEM) using high-resolution satellite images requires rigorous sensor modelling or highly accurate approximations to the rigorous sensor model, such as rational polynomial coefficients (RPCs) including methods for the removal of exterior orientation biases [1,2,3]
The KOMPSAT-3 images cover an area of approximTahteeltyes2t0siktemfobryth1e6exkpmeriinmethnetsYisasnhgoswann icnitFyigruergeio4.nT. hTehKeOvMerPySAmTo-u3 nimtaaigneosucsovaerreaanisarfeoareosfted at higheahrpigaphlrteoirtxuaimdlteiatstue.dlyTeas2.b0Tlkeamb1lesb1hyos1hw6okswmtshtiehnedthdeeetatYailialsnsogosffattnhheeciptpyaanrncechghrioornom.maTtahicteiKcvOeKrMyOPmMSoAPuTSnA-t3aTiinm-3oauigmseaaprgeaaier.ipsTafhoierr.egTsrtoheudenagdtround resoluretsioolnutoiofnthoef tthweotwimo aimgeagseiss i8s585cmcmaanndd8866 ccmm, rreessppeecctitviveleyl.yR. oRlloallndanydawyaawngalensgolfebsoothf bimotahgeismages pointpioninatsiinmaislaimr idlairredcitrieocnti.onT.hTehpeiptcithchaanngglelessddiiffffeerr eesssseenntitaialllylyininthtehiresirigsni.gn
This paper proposes a point-to-surface matching procedure and applies it to ground control points (GCPs) and a DEM which is generated from KOMPSAT-3 satellite images with vendor-provided RPCs

Summary

Introduction

The generation of a high-quality digital elevation model (DEM) using high-resolution satellite images requires rigorous sensor modelling or highly accurate approximations to the rigorous sensor model, such as rational polynomial coefficients (RPCs) including methods for the removal of exterior orientation biases [1,2,3]. For the KOMPSAT-3’s images and RPCs that are used in the experimental investigations, the absolute accuracy specification is approximately 50 m with circular error at 90% probability (CE90) [4]. As demonstrated for many high-resolution imaging satellites using RPC georeferencing, there is ample experimental evidence for KOMPSAT imagery that relative accuracy to meter level is attainable. The GCPs have to be measured in the images to make them usable for the methods

Objectives

Results

Conclusion