Bridge Modeling and Application for Determining In-Flight Sensor Spatial Resolution

In-fight estimation of sensor spatial performance can be assessed at various ground targets. When assessing sensors of pixel sizes ~10 m - 100m, bridges over water are a common target site. The bridge acts as a linear impulse signal, and a line spread function (LSF) can be directly extrapolated from the image over a bridge. One point of interference in this extracted LSF is artificial widening caused by the bridge width. Even though it is treated as an impulse, if the bridge is too wide for the pixel size it can smear the LSF. This would make the imagery appear to be of lower quality than it really is. To remove the impacts of bridge width, we run a series of simulations where a modeled perfect bridge has a known gaussian LSF applied, outputting a simulated image of the bridge. The LSF is estimated from the output image as it would be in practice with a real image, and the apparent widened Full Width at Half Maximum (FWHM) recorded. This is performed on various combinations of bridge width relative to pixel size, to build a non-dimensional empirical relationship between apparent FWHM and true FWHM. We find that if the bridge width is 0.2 times the pixel size or less, the apparent FWHM is equal to the true FWHM. For larger ratios, this empirical relationship can be applied to any imaged bridge, allowing for easy removal of the impacts of bridge width. Proof of concept is performed with Landsat-8/9 and Sentinel-2.