Fluorescence-based classification of Caribbean coral reef organisms and substrates.

A diverse group of coral reef organisms, representing several phyla, possess fluorescent pigments. We investigated the potential of using the characteristic fluorescence emission spectra of these pigments to enable unsupervised, optical classification of coral reef habitats. We compiled a library of...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: David G Zawada, Charles H Mazel
Formato: article
Lenguaje:EN
Publicado: Public Library of Science (PLoS) 2014
Materias:
R
Q
Acceso en línea:https://doaj.org/article/210556027b5d42c6bbb59f3be30f4a03
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
id oai:doaj.org-article:210556027b5d42c6bbb59f3be30f4a03
record_format dspace
spelling oai:doaj.org-article:210556027b5d42c6bbb59f3be30f4a032021-11-18T08:37:34ZFluorescence-based classification of Caribbean coral reef organisms and substrates.1932-620310.1371/journal.pone.0084570https://doaj.org/article/210556027b5d42c6bbb59f3be30f4a032014-01-01T00:00:00Zhttps://www.ncbi.nlm.nih.gov/pmc/articles/pmid/24482676/pdf/?tool=EBIhttps://doaj.org/toc/1932-6203A diverse group of coral reef organisms, representing several phyla, possess fluorescent pigments. We investigated the potential of using the characteristic fluorescence emission spectra of these pigments to enable unsupervised, optical classification of coral reef habitats. We compiled a library of characteristic fluorescence spectra through in situ and laboratory measurements from a variety of specimens throughout the Caribbean. Because fluorescent pigments are not species-specific, the spectral library is organized in terms of 15 functional groups. We investigated the spectral separability of the functional groups in terms of the number of wavebands required to distinguish between them, using the similarity measures Spectral Angle Mapper (SAM), Spectral Information Divergence (SID), SID-SAM mixed measure, and Mahalanobis distance. This set of measures represents geometric, stochastic, joint geometric-stochastic, and statistical approaches to classifying spectra. Our hyperspectral fluorescence data were used to generate sets of 4-, 6-, and 8-waveband spectra, including random variations in relative signal amplitude, spectral peak shifts, and water-column attenuation. Each set consisted of 2 different band definitions: 'optimally-picked' and 'evenly-spaced.' The optimally-picked wavebands were chosen to coincide with as many peaks as possible in the functional group spectra. Reference libraries were formed from half of the spectra in each set and used for training purposes. Average classification accuracies ranged from 76.3% for SAM with 4 evenly-spaced wavebands to 93.8% for Mahalanobis distance with 8 evenly-spaced wavebands. The Mahalanobis distance consistently outperformed the other measures. In a second test, empirically-measured spectra were classified using the same reference libraries and the Mahalanobis distance for just the 8 evenly-spaced waveband case. Average classification accuracies were 84% and 87%, corresponding to the extremes in modeled water-column attenuation. The classification results from both tests indicate that a high degree of separability among the 15 fluorescent-spectra functional groups is possible using only a modest number of spectral bands.David G ZawadaCharles H MazelPublic Library of Science (PLoS)articleMedicineRScienceQENPLoS ONE, Vol 9, Iss 1, p e84570 (2014)
institution DOAJ
collection DOAJ
language EN
topic Medicine
R
Science
Q
spellingShingle Medicine
R
Science
Q
David G Zawada
Charles H Mazel
Fluorescence-based classification of Caribbean coral reef organisms and substrates.
description A diverse group of coral reef organisms, representing several phyla, possess fluorescent pigments. We investigated the potential of using the characteristic fluorescence emission spectra of these pigments to enable unsupervised, optical classification of coral reef habitats. We compiled a library of characteristic fluorescence spectra through in situ and laboratory measurements from a variety of specimens throughout the Caribbean. Because fluorescent pigments are not species-specific, the spectral library is organized in terms of 15 functional groups. We investigated the spectral separability of the functional groups in terms of the number of wavebands required to distinguish between them, using the similarity measures Spectral Angle Mapper (SAM), Spectral Information Divergence (SID), SID-SAM mixed measure, and Mahalanobis distance. This set of measures represents geometric, stochastic, joint geometric-stochastic, and statistical approaches to classifying spectra. Our hyperspectral fluorescence data were used to generate sets of 4-, 6-, and 8-waveband spectra, including random variations in relative signal amplitude, spectral peak shifts, and water-column attenuation. Each set consisted of 2 different band definitions: 'optimally-picked' and 'evenly-spaced.' The optimally-picked wavebands were chosen to coincide with as many peaks as possible in the functional group spectra. Reference libraries were formed from half of the spectra in each set and used for training purposes. Average classification accuracies ranged from 76.3% for SAM with 4 evenly-spaced wavebands to 93.8% for Mahalanobis distance with 8 evenly-spaced wavebands. The Mahalanobis distance consistently outperformed the other measures. In a second test, empirically-measured spectra were classified using the same reference libraries and the Mahalanobis distance for just the 8 evenly-spaced waveband case. Average classification accuracies were 84% and 87%, corresponding to the extremes in modeled water-column attenuation. The classification results from both tests indicate that a high degree of separability among the 15 fluorescent-spectra functional groups is possible using only a modest number of spectral bands.
format article
author David G Zawada
Charles H Mazel
author_facet David G Zawada
Charles H Mazel
author_sort David G Zawada
title Fluorescence-based classification of Caribbean coral reef organisms and substrates.
title_short Fluorescence-based classification of Caribbean coral reef organisms and substrates.
title_full Fluorescence-based classification of Caribbean coral reef organisms and substrates.
title_fullStr Fluorescence-based classification of Caribbean coral reef organisms and substrates.
title_full_unstemmed Fluorescence-based classification of Caribbean coral reef organisms and substrates.
title_sort fluorescence-based classification of caribbean coral reef organisms and substrates.
publisher Public Library of Science (PLoS)
publishDate 2014
url https://doaj.org/article/210556027b5d42c6bbb59f3be30f4a03
work_keys_str_mv AT davidgzawada fluorescencebasedclassificationofcaribbeancoralreeforganismsandsubstrates
AT charleshmazel fluorescencebasedclassificationofcaribbeancoralreeforganismsandsubstrates
_version_ 1718421548562382848