Empirical correction of multiple flightline hyperspectral aerial image mosaics

Aerial survey provides the user with great flexibility in terms of the geometry of sensing and the timing of measurements, but mosaicking individual aerial images to produce an extensive coverage remains a problem. Empirical methods based on normalising individual images to a common standard are used widely to create visually acceptable mosaics. However, the effect of these methods on quantitative estimation of land surface properties is unknown. An existing method for atmospherically correcting an aerial image mosaic involves fitting a regression model using pixels from the overlapping edges of adjacent flightlines. Here, we demonstrate a new method of atmospherically correcting an aerial image mosaic, based on use of an additional orthogonal flightline. The two methods were compared by using the two image mosaics to calculate vegetation indices (NDVI, SAVI, ARVI), which were then used to predict leaf area index, which was known in detail from ground survey. The second method was found to have lower uncertainty for all three vegetation indices tested. ARVI was found to be the most robust of the three when applied across multiple flightlines, regardless of the method of atmospheric correction.

Research Highlights
► LAI estimation is improved by using data from an orthogonal flightline.
► Robust regression is preferred for this task.
► ARVI is less affected by between-flightline variations than NDVI or SAVI.