Effects of signature mismatch on hyperspectral detection algorithms

The main objective of this paper is to discuss the effects of signature mismatch on hyperspectral target detection algorithms. The main causes of mismatch are atmospheric propagation, intrinsic spectral variability, sensor noise, and sensor artifacts. We provide a theoretical analysis that shows the effects of mismatch on adaptive detection algorithms, which use estimates of background covariance matrix, and we present a systematic diagonal loading technique which provides controlled robustness to mismatch.