Tapered-shrinkage covariance matrix estimation and its application to hyperspectral target detection

Le résumé fourni par la source

Hyperspectral target detection mainly relies on an accurate covariance matrix estimation, particularly in high-dimensional settings. Tapering-based covariance estimators have improved such estimation in many applications. In this paper, we investigate the impact of different tapering weights within the TApered or BAnded Shrinkage COvariance matrix (TABASCO) framework. We demonstrate that both the exponential and wendland non-linear tapering weights can significantly improve the target detection performance in both Monte-Carlo simulations and hyperspectral data.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.