File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Using CASI hyperspectral imagery to detect mortality and vegetation stress associated with a new hardwood forest disease

TitleUsing CASI hyperspectral imagery to detect mortality and vegetation stress associated with a new hardwood forest disease
Authors
Issue Date2008
Citation
Photogrammetric Engineering and Remote Sensing, 2008, v. 74, n. 1, p. 65-75 How to Cite?
AbstractA Compact Airborne Spectrographic Imager-2 (CASI) dataset was used for detecting mortality and vegetation stress associated with a new forest disease. We first developed a multilevel classification scheme to improve classification accuracy. Then, the CASI raw data were transformed to reflectance and corrected for topography, and a principal component (PC) transformation of all 48 bands and the visible bands and NIR bands were separately conducted to extract features from the CASI data. Finally, we classified the calibrated and corrected CASI imagery using a maximum likelihood classifier and tested the relative accuracies of classification across the scheme. The multilevel scheme consists of four levels (Levels 0 to 3). Level 0 covered the entire study area, classifying eight classes (oak trees, California bay trees, shrub areas, grasses, dead trees, dry areas, wet areas, and water). At Level 1, the vegetated and non-vegetated areas were separated. The vegetated and non-vegetated areas were further subdivided into four vegetated (oak trees, California bay trees, shrub areas, grasses) and four non-vegetated (dead trees, dry areas, wet areas, and water) classes at Level 2. Level 3 identified stressed and non-stressed oak trees (two classes). The ten classes classified at different levels are defined as final classes in this study. The experimental results indicated that classification accuracy generally increased as the detailed classification level increased. When the CASI topographically corrected reflectance data were processed into ten PCs (five PCs from the visible region and five PCs from NIR bands), the classification accuracy for Level 2 vegetated classes (non-vegetated classes) increased to 80.15 percent (94.10 percent) from 78.07 percent (92.66 percent) at Level 0. The accuracy of separating stressed from non-stressed oak trees at Level 3 was 75.55 percent. When classified as a part of Level 0, the stressed and non-stressed were almost inseparable. Furthermore, we found that PCs derived from visible and NIR bands separately yielded more accurate results than the PCs from all 48 CASI bands. © 2008 American Society for Photogrammetry and Remote Sensing.
Persistent Identifierhttp://hdl.handle.net/10722/296618
ISSN
2023 Impact Factor: 1.0
2023 SCImago Journal Rankings: 0.309
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorPu, Ruiliang-
dc.contributor.authorKelly, Maggi-
dc.contributor.authorAnderson, Gerald L.-
dc.contributor.authorGong, Peng-
dc.date.accessioned2021-02-25T15:16:17Z-
dc.date.available2021-02-25T15:16:17Z-
dc.date.issued2008-
dc.identifier.citationPhotogrammetric Engineering and Remote Sensing, 2008, v. 74, n. 1, p. 65-75-
dc.identifier.issn0099-1112-
dc.identifier.urihttp://hdl.handle.net/10722/296618-
dc.description.abstractA Compact Airborne Spectrographic Imager-2 (CASI) dataset was used for detecting mortality and vegetation stress associated with a new forest disease. We first developed a multilevel classification scheme to improve classification accuracy. Then, the CASI raw data were transformed to reflectance and corrected for topography, and a principal component (PC) transformation of all 48 bands and the visible bands and NIR bands were separately conducted to extract features from the CASI data. Finally, we classified the calibrated and corrected CASI imagery using a maximum likelihood classifier and tested the relative accuracies of classification across the scheme. The multilevel scheme consists of four levels (Levels 0 to 3). Level 0 covered the entire study area, classifying eight classes (oak trees, California bay trees, shrub areas, grasses, dead trees, dry areas, wet areas, and water). At Level 1, the vegetated and non-vegetated areas were separated. The vegetated and non-vegetated areas were further subdivided into four vegetated (oak trees, California bay trees, shrub areas, grasses) and four non-vegetated (dead trees, dry areas, wet areas, and water) classes at Level 2. Level 3 identified stressed and non-stressed oak trees (two classes). The ten classes classified at different levels are defined as final classes in this study. The experimental results indicated that classification accuracy generally increased as the detailed classification level increased. When the CASI topographically corrected reflectance data were processed into ten PCs (five PCs from the visible region and five PCs from NIR bands), the classification accuracy for Level 2 vegetated classes (non-vegetated classes) increased to 80.15 percent (94.10 percent) from 78.07 percent (92.66 percent) at Level 0. The accuracy of separating stressed from non-stressed oak trees at Level 3 was 75.55 percent. When classified as a part of Level 0, the stressed and non-stressed were almost inseparable. Furthermore, we found that PCs derived from visible and NIR bands separately yielded more accurate results than the PCs from all 48 CASI bands. © 2008 American Society for Photogrammetry and Remote Sensing.-
dc.languageeng-
dc.relation.ispartofPhotogrammetric Engineering and Remote Sensing-
dc.titleUsing CASI hyperspectral imagery to detect mortality and vegetation stress associated with a new hardwood forest disease-
dc.typeArticle-
dc.description.naturelink_to_OA_fulltext-
dc.identifier.doi10.14358/PERS.74.1.65-
dc.identifier.scopuseid_2-s2.0-37749014733-
dc.identifier.volume74-
dc.identifier.issue1-
dc.identifier.spage65-
dc.identifier.epage75-
dc.identifier.isiWOS:000252077700008-
dc.identifier.issnl0099-1112-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats