Presenter Information

Minsu Kim, KBRwyle

Abstract

A conventional method to evaluate the spatial performance of a small pixel satellite image uses a high contrast edge target. Because the largest manageable cal/val site is about 50 meter in length along the edge, the GSD of the applicable satellite image is about 2-meter or less in case of a desirable edge orientation relative to satellite orbit path. Thus, to overcome the lack of cal/val grade edge target for a larger GSD image, an alternative method is to utilize large farmland with an interface by two different crop types or different cultivation conditions. However, it is very hard to find a farmland target that satisfies the requirement, which is the two large homogenous areas with high contrast along a perfect straight line interface. Also, the lack of controls on the crop growth stage and cultivation condition calls for other techniques.

We present a method to evaluate the spatial performance of a satellite image with moderately large GSD (roughly a few meters to thirty or so meters) using a bridge-like features over optically deep water background. The technique begins with an analog bridge modeling using several parameters. The minimum required parameters for analog modeling are the angle of the bridge orientation relative to the satellite orbit and the width of the bridge. In case of a twin bridge, the length of a gap between two bridges is also critical. The different width between the two individual bridge of the twin bridge are also allowed in the analog modeling. The arbitrary traffic condition on the bridge is assumed as a random radiometric variable. The next step is the discrete sampling of the analog bridge model using the GSD and the PSF (impulse response function) of the sensor in the satellite orbit coordinate. Then, the discrete sampled image in the satellite coordinate is resampled to the image in the map projected coordinate, which is the final image provided to the public. The method to determine the FWHM of the Gaussian PSF is to use an optimization that iteratively changes the FWHM until the difference between the projected image of a given FWHM and the bridge image segment of the satellite image is minimized. The FWHM of the PSF of a sensor is determined from multiple bridges of varying physical dimensions and orientations. We present the spatial analysis result for Landsat OLI, Sentinel-2 MSI, and the EnMAP sensors.

Share

COinS
 
Jun 11th, 4:05 PM

Spatial Performance Analysis of Satellite Images using Causeways

A conventional method to evaluate the spatial performance of a small pixel satellite image uses a high contrast edge target. Because the largest manageable cal/val site is about 50 meter in length along the edge, the GSD of the applicable satellite image is about 2-meter or less in case of a desirable edge orientation relative to satellite orbit path. Thus, to overcome the lack of cal/val grade edge target for a larger GSD image, an alternative method is to utilize large farmland with an interface by two different crop types or different cultivation conditions. However, it is very hard to find a farmland target that satisfies the requirement, which is the two large homogenous areas with high contrast along a perfect straight line interface. Also, the lack of controls on the crop growth stage and cultivation condition calls for other techniques.

We present a method to evaluate the spatial performance of a satellite image with moderately large GSD (roughly a few meters to thirty or so meters) using a bridge-like features over optically deep water background. The technique begins with an analog bridge modeling using several parameters. The minimum required parameters for analog modeling are the angle of the bridge orientation relative to the satellite orbit and the width of the bridge. In case of a twin bridge, the length of a gap between two bridges is also critical. The different width between the two individual bridge of the twin bridge are also allowed in the analog modeling. The arbitrary traffic condition on the bridge is assumed as a random radiometric variable. The next step is the discrete sampling of the analog bridge model using the GSD and the PSF (impulse response function) of the sensor in the satellite orbit coordinate. Then, the discrete sampled image in the satellite coordinate is resampled to the image in the map projected coordinate, which is the final image provided to the public. The method to determine the FWHM of the Gaussian PSF is to use an optimization that iteratively changes the FWHM until the difference between the projected image of a given FWHM and the bridge image segment of the satellite image is minimized. The FWHM of the PSF of a sensor is determined from multiple bridges of varying physical dimensions and orientations. We present the spatial analysis result for Landsat OLI, Sentinel-2 MSI, and the EnMAP sensors.