I am attempting to use ERDAS target detection models to detect/differentiate a specific type of soil from a general background. We have gathered spectral signatures of the target soil in the field with a spectrometer.
The target soil has only slight, but very consistent, variation from the prevalent background signatures.
We have attempted to run target detection using the field derived spectrum on a FLAASH corrected Hyperion image.
However, the FLAASH correction does not appear smooth enough to allow for the slight variation between target and background signatures to be observed.
We are considering using the atmospheric adjustment tool (Modified Flat Field approach) to improve our initial FLAASH atmospheric correction.
We would like to use our target signature as the reference for this atmospheric adjustment, so that the resulting adjusted image correctly displays the key spectral profile features of the target.
Would doing such an adjustment (using the same signature to adjust and then detect) be a circular process and therefore not viable? We believe this process may give us the best chance at detection by enhancing the accuracy around the subtle spectral features of our target.
We are concerned that the workflow might not be considered valid, or be seen as a unfair bias influencing our detection.