Target Detection in Airborne Hyperspectral Imagery and its Sensitivity to Different Atmospheric Correction Methods

Fine resolution hyperspectral imagery has promising application in the detection of natural and synthesized objects. Atmospheric correction (AC) is an important part of the pre-processing chain involved in target detection from remote sensing imagery. Different AC model approximates the atmosphere using different mathematical assumptions, resulting in different quality of the surface reflectance obtained. However, the effects of the choice of AC models on the overall application of hyperspectral imagery for target identification is unaddressed. We designed an experimental setup at Gudalur town, Tamil Nadu, India to study the effects of AC models on the performance of target identification using airborne hyperspectral imagery. State-of the-art spectral unmixing and target detection algorithms are applied to hyperspectral imagery for performance assessment. Results indicate a substantial aberration of target detection performance as a function of the atmospheric correction model.