Authors / CoAuthors
Tan, P. | Lymburner, L. | Mueller, N.
Abstract
Abstract: In most cases a single pixel in a satellite image contains information from more than one type of land cover substance. One challenge is to decompose a pixel with mixed spectral readings into a set of endmembers, and estimate the corresponding abundance fractions. The linear spectral unmixing model assumes that spectral reading of a single pixel is a linear combination of spectral readings from a set of endmembers. Most linear spectral unmixing algorithms rely on spectral signatures from endmembers in pre-defined libraries obtained from previous on-ground studies. Therefore, the applications of these algorithms are restricted to images whose extent and acquisition time coincide with those of the endmember library. We propose a linear spectral unmixing algorithm which is able to identify a set of endmembers from the actual image of the studied area. Existing spectral libraries are used as training sets to infer a model which determines the class labels of the derived image based endmembers. The advantage of such an approach is that it is capable of performing consistent spectral unmixing in areas with no established endmember libraries. Testing has been conducted on a Landsat7 ETM+ image subset of the Gwydir region acquired on Jun 22, 2008. Three types of land cover classes: bare soil, green vegetation and non-photosynthetic are specified for this test. A set consisting of 150 endmember samples and a number of ground abundance observations were obtained from a corresponding field trip. The study successfully identified an endmember set from the image for the specified land cover classes. For most test points, the spectral unmixing and estimation of the corresponding abundance are consistent with the ground validation data. From the 20th International Congress on Modelling and Simulation (MODSIM2013)
Product Type
nonGeographicDataset
eCat Id
77026
Contact for the resource
Custodian
Point of contact
Cnr Jerrabomberra Ave and Hindmarsh Dr GPO Box 378
Canberra
ACT
2601
Australia
Keywords
-
- External PublicationConference Paper
- Australian and New Zealand Standard Research Classification (ANZSRC)
-
- Earth Sciences
-
- Published_Internal
Publication Date
2013-01-01T00:00:00
Creation Date
Security Constraints
Legal Constraints
Status
Purpose
Maintenance Information
unknown
Topic Category
geoscientificInformation
Series Information
Lineage
Unknown
Parent Information
Extents
Reference System
Spatial Resolution
Service Information
Associations
Downloads and Links
Source Information
Source data not available.