Created with Sketch.
Validation of Forward Modeling Target Detection Approach on a New Hysperpectral Data Set Featuring an Urban Scenario and Variable Illumination Conditions
Target detection in hyperspectral imagery has been shown to be a powerful tool to search for specific objects or materials (i.e., the targets) in a remotely sensed scene. By exploiting knowledge of the spectral signature of the target of interest, the goal is to identify pixels in the image whose spectra exhibit a high similarity degree with the target spectral signature. When the hyperspectral image is acquired by a sensor operating in the visible, near infrared, and short-wave infrared (VNIR-SWIR) spectral ranges – where reflection phenomena dominate– the spectral reflectance of the target serves as its spectral signature. Since hyperspectral sensors acquire imagery data in terms of spectral radiance, radiation transfer (RT) in the atmosphere needs to be accounted for in order to properly accomplish the target detection task. One effective approach is the so-called forward modeling (FM) approach, originally conceived by Healey and Slater. FM basically consists of transforming the target reflectance spectral signature into a target spectral radiance by accounting for a physical model for the atmospheric RT and by specifying the model parameters on the basis of atmospheric, viewing, and illumination conditions at acquisition time. The FM approach has been revealed particularly useful when there is the need to allow for a certain degree of flexibility with respect to variability in the acquisition conditions, such as when atmospheric/illumination conditions are not accurately known or when they are spatially variable in the scene. Goal of this work is validating the FM-based target detection approach on a new hyperspectral data set featuring several different targets deployed in an urban scenario characterized by variable and non-accurately known acquisition conditions.
Acito Nicola, Corsini Giovanni, Diani Marco, Matteoli Stefania, Riccobono Aldo, Rossi Alessandro
Paper for Seminar/Symposium/Conference
IGARSS 2015 - IEEE International Geoscience and Remote Sensing Symposium (26-31 July 2015, Milan, Italy)
Inserisci il testo di ricerca