A comparison of real and simulated airborne hyperspectral imagery

Simulated remote sensing imagery can be very valuable for instrument design studies, analysis algorithm development and validation, and in verification of real image phenomenology. However, the question of simulation model validity and verification is important. This paper contributes to this topic by presenting a comprehensive comparison methodology between real and simulated hyperspectral imagery. SpecTIR acquired a real hyperspectral image including 360 bands spanning from 400 to 2450 nm, while the simulated one was generated with DIRSIG, an image generation tool developed by the Rochester Institute of Technology. This paper analyzes the image covariance matrix eigenstructure, and presents results for classification, unmixing and sub-pixel target detection. The comparison demonstrates that the simulated image captures the dominant characteristics of the real image.