In this paper we deal with the full pixel target detection problem in hyperspectral images. We review the statistical approach to the problem and we derive the algorithms based on the replacement target model. This model characterises the interactions between the target and the background. In this framework we derive three algorithms: the Quadratic Detector (QD), the Adaptive Detector (AD) and the Adaptive Cosine Estimator (ACE). For each algorithm we discuss the limits related to their use in actual applications. We make use of an experimental dataset acquired by the MIVIS sensor in order to compare the performance of the AD and the Adaptive Matched Filter under the replacement target hypothesis and to verify that the ACE has not the CFAR property
Full-Pixel Target Detection in Hyperspectral Data: a Review
Acito N;CORSINI, GIOVANNI;DIANI, MARCO
2003-01-01
Abstract
In this paper we deal with the full pixel target detection problem in hyperspectral images. We review the statistical approach to the problem and we derive the algorithms based on the replacement target model. This model characterises the interactions between the target and the background. In this framework we derive three algorithms: the Quadratic Detector (QD), the Adaptive Detector (AD) and the Adaptive Cosine Estimator (ACE). For each algorithm we discuss the limits related to their use in actual applications. We make use of an experimental dataset acquired by the MIVIS sensor in order to compare the performance of the AD and the Adaptive Matched Filter under the replacement target hypothesis and to verify that the ACE has not the CFAR propertyI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
