09:00 〜 10:30
[ACG37-P17] Geolocation accuracy assessment of ICEYE microsatellite images
キーワード：ICEYE, SAR image, geolocation accuracy
The ICEYE is a constellation of X-band Synthetic Aperture Radar (SAR) Earth observation microsatellites that offer submeter resolution images. These high-resolution SAR images have various uses, including disaster monitoring and displacement mapping. The ICEYE’s image product includes Ground Range Detected (GRD) and Single Look Complex (SLC) image. The GRD images show amplitude only, while SLC images include both amplitude and phase information.
Geometric correction is a crucial step in the preprocessing of SAR images. After this process, the SAR data can be combined with other geospatial information for analysis and comparison. ICEYE's geometric correction can be performed by precise orbital parameters or Rational Polynomial Coefficients (RPC). The precise orbital parameters are typically used in SLC images for Interferometric Synthetic Aperture Radar (InSAR) applications, while the RPC is often used in GRD images for mapping.
This study aims to perform geolocation accuracy assessment for ICEYE’s GRD satellite images. Moreover, the GRD images in ascending and descending modes are also compared in accuracy analysis.
This study uses Rational Function Models (RFM) built from the RPCs in image metadata. The RFM describes the relationship between world and image coordinates using a ratio of polynomial functions. The geometric correction is performed using several ground control points to correct for geolocation errors, and independent check points are used to assess the accuracy of the geolocation.
The accuracy of geolocation in 11 ICEYE GRD images in Taiwan was evaluated using ground control points and independent check points from Taiwan eMap with an accuracy of 1.25m. The elevation of points was measured from ALOS World 3D-30m Digital Surface Model. The mean RMSE for the 11 images was 6.7 pixels in range direction and 2.0 pixels in azimuth direction, with the accuracy being better in azimuth direction. The mean RMSE for descending images were 6.3 pixels and 1.9 pixels. The mean RMSE for ascending images were 7.0 pixels and 2.1 pixels. The mean RMSE for descending images was slightly better than for ascending images. In summary, the geolocation accuracy of ICEYE GRD images for Earth environment observation was better than 7 pixels in range direction and 2 pixels in azimuth direction.
Geometric correction is a crucial step in the preprocessing of SAR images. After this process, the SAR data can be combined with other geospatial information for analysis and comparison. ICEYE's geometric correction can be performed by precise orbital parameters or Rational Polynomial Coefficients (RPC). The precise orbital parameters are typically used in SLC images for Interferometric Synthetic Aperture Radar (InSAR) applications, while the RPC is often used in GRD images for mapping.
This study aims to perform geolocation accuracy assessment for ICEYE’s GRD satellite images. Moreover, the GRD images in ascending and descending modes are also compared in accuracy analysis.
This study uses Rational Function Models (RFM) built from the RPCs in image metadata. The RFM describes the relationship between world and image coordinates using a ratio of polynomial functions. The geometric correction is performed using several ground control points to correct for geolocation errors, and independent check points are used to assess the accuracy of the geolocation.
The accuracy of geolocation in 11 ICEYE GRD images in Taiwan was evaluated using ground control points and independent check points from Taiwan eMap with an accuracy of 1.25m. The elevation of points was measured from ALOS World 3D-30m Digital Surface Model. The mean RMSE for the 11 images was 6.7 pixels in range direction and 2.0 pixels in azimuth direction, with the accuracy being better in azimuth direction. The mean RMSE for descending images were 6.3 pixels and 1.9 pixels. The mean RMSE for ascending images were 7.0 pixels and 2.1 pixels. The mean RMSE for descending images was slightly better than for ascending images. In summary, the geolocation accuracy of ICEYE GRD images for Earth environment observation was better than 7 pixels in range direction and 2 pixels in azimuth direction.