B Bahrambeygi; H Ranjbar; J Shahabpour; S.H Moeinzadeh
Abstract
Hyperspectral images of Hyperion sensor is a rich source of information with 242 narrow contiguous spectral bands. Among these, there are a number of atmospheric agents, which contaminate the content of various information bands. Therefore, to obtain the complete advantage of a Hyperspectral image in ...
Read More
Hyperspectral images of Hyperion sensor is a rich source of information with 242 narrow contiguous spectral bands. Among these, there are a number of atmospheric agents, which contaminate the content of various information bands. Therefore, to obtain the complete advantage of a Hyperspectral image in optimum condition, atmospheric correction is an inevitable process. Atmospheric corrections may be conducted by two methods, namely data based, and scene based. In the scene based methods, spectral anomalies are detected and corrected by using self-image spectral information processing without a field information requirement. In this study, two scene based atmospheric correction methods of Quick Atmospheric Correction (QUAC) and Internal Average Relative Reflectance (IARR) were examined on Hyperion image of Masahim volcanic crater. To evaluate the results of these two scenesbased methods, the results of field spectroscopy and data based empirical line method were used. X-ray diffraction and spectral analysis of selected samples, whose locations were determined through SAM method, illustrated kaolinite pattern as index mineral of argillic zone.In order to compare the results obtained from different atmospherically corrected images quantitatively, maximum probability pixels obtained from SAM method were evaluated for each corrected images in classified information format. After drawing the accuracy matrix for classified pixels and sampled and investigated pixels in the field and laboratory studies, the accuracy coefficients were calculated for the favorable districts of the corrected images bytwo scene based methods and ELM method. The examination results display the producer accuracy of 74.58 percent for IARR corrected images and a producer accuracy of 35.5 percent for QUAC corrected image; whereas the ELM data based correction method despite using field spectrometry data shows the producer accuracy of 74.58 percent. Therefore, in discrimination of argillic zone in semi-arid regions, IARR atmospheric correction method is considered as suitable and affordable preprocessing method to retrieve spectral information from the hyperspectral data.
