TY - JOUR
T1 - PCA-based classification using airborne hyperspectral radiance data, a case study
T2 - Mount Horshan Mediterranean forest
AU - Mandelmilch, Moshe
AU - Dadon, Alon
AU - Ben-Dor, Eyal
AU - Sheffer, Efrat
N1 - Publisher Copyright:
© 2021 Informa UK Limited, trading as Taylor & Francis Group.
PY - 2022
Y1 - 2022
N2 - Atmospheric correction (ATC) of radiance image data is a preliminary and necessary procedure to reach a coherent unsupervised classification. Though ATC results in removal of noise artefacts related to path radiance, loss of some data is inherent by the process. The unsupervised principal component analysis-based classification (PCABC) was harnessed in this paper using radiance data that bypass the ATC protocol. Being primarily based on the variability of the input hyperspectral remote sensing (HRS) image regardless of its physical attributes, it was assumed that PCABC can be applied to radiance HRS image just as already shown on reflectance domain. To test this assumption, PCABC was tested on a radiance HRS image of Specim’s AisaFENIX taken over the Mediterranean forest of Mount Horshan, Israel. With no application of ATC or noise reduction, while tested unsupervised classification methods were insufficient, PCABC was able to classify four different plant species with an overall accuracy of 68%.
AB - Atmospheric correction (ATC) of radiance image data is a preliminary and necessary procedure to reach a coherent unsupervised classification. Though ATC results in removal of noise artefacts related to path radiance, loss of some data is inherent by the process. The unsupervised principal component analysis-based classification (PCABC) was harnessed in this paper using radiance data that bypass the ATC protocol. Being primarily based on the variability of the input hyperspectral remote sensing (HRS) image regardless of its physical attributes, it was assumed that PCABC can be applied to radiance HRS image just as already shown on reflectance domain. To test this assumption, PCABC was tested on a radiance HRS image of Specim’s AisaFENIX taken over the Mediterranean forest of Mount Horshan, Israel. With no application of ATC or noise reduction, while tested unsupervised classification methods were insufficient, PCABC was able to classify four different plant species with an overall accuracy of 68%.
KW - Hyperspectral remote sensing
KW - plant species classification
KW - radiance image
KW - unsupervised classification
UR - http://www.scopus.com/inward/record.url?scp=85108182293&partnerID=8YFLogxK
U2 - 10.1080/10106049.2021.1923830
DO - 10.1080/10106049.2021.1923830
M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???
AN - SCOPUS:85108182293
SN - 1010-6049
VL - 37
SP - 5783
EP - 5806
JO - Geocarto International
JF - Geocarto International
IS - 20
ER -