Geometric Accuracy Analysis of Geostationary Environment Monitoring Spectrometer Images by Spatiotemporal and Spectral Collocation

초록

Geometric accuracy of satellite imagery is a critical factor that determines the reliability of image analysis. Geostationary Environment Monitoring Spectrometer (GEMS) is the world's first geostationary hyperspectral environmental satellite launched in 2020. Unique geometric and radiometric characteristics of GEMS make conventional techniques difficult to be applied for automated geometric accuracy evaluation of GEMS images. This study aims to handle this problem by matching GEMS images with images from Advanced Meteorological Imager (AMI), a meteorological geostationary satellite. Since GEMS and AMI differ in scan patterns and spectral responses, their spatiotemporal and spectral characteristics are different. For accurate matching, we apply collocation techniques to make these characteristics conform to each other. Geometric accuracy was evaluated through global and local matching. Global matching results showed that geometric accuracy of GEMS images exhibited a repetitive error trend of about 1 pixel in the north–south direction over time. Local matching enabled high-density analysis of detailed error distribution and direction across the entire image. Furthermore, comparison of results with and without collocation revealed that excluding spectral collocation increased irregular errors across the image. Without strict temporal collocation, geometric errors grew in regions strongly affected by observation time differences. In contrast, when spatial, spectral, and strict temporal collocation were all applied, the average root mean square error remained stable at about 1–1.5 pixels. These findings demonstrate that the collocation process is an essential preprocessing step for evaluating geometric accuracy of geostationary hyperspectral satellites such as GEMS. Future research will expand toward developing algorithms to improve geometric accuracy of the GEMS based on these results. © 2008-2012 IEEE.

키워드