Skip to main content

page search

Library Mapping impervious surfaces from superresolution enhanced CHRIS/Proba imagery using multiple endmember unmixing

Mapping impervious surfaces from superresolution enhanced CHRIS/Proba imagery using multiple endmember unmixing

Mapping impervious surfaces from superresolution enhanced CHRIS/Proba imagery using multiple endmember unmixing

Resource information

Date of publication
December 2012
Resource Language
ISBN / Resource ID
AGRIS:US201600059922
Pages
99-112

In this paper, the potential of superresolution (SR) image reconstruction methods for sub-pixel land-cover mapping in dense urban areas is studied. A multiple endmember approach (MESMA) is used for unmixing both original hyperspectral CHRIS/Proba and SR enhanced CHRIS/Proba data. Validation based on high resolution orthophotos (25cm) shows that land-cover fraction maps generated from SR-enhanced CHRIS/Proba data (9m) have a lower overall fractional error compared to the land-cover fractions produced from the original CHRIS data (18m), when validating both results at the 18m resolution. Validation of SR results at the 9m resolution produces an overall mean absolute error (OMAE) of 16.7% compared to an OMAE of 14.3% at the 18m resolution, with the original data, yet the impervious surface map produced at 9m has a much higher level of detail than the original map, better representing the built-up pattern of the urban environment. Detailed analysis of impervious surface mapping results for different reference proportion intervals points at smaller average fractional errors for impervious surface fractions produced from SR-enhanced data when validation is done at the original 18m resolution, over the entire range of proportions. Only for pixels not containing any impervious surface cover the fractional error is marginally higher than with the original data. These results demonstrate the potential of SR-enhanced data for more accurate impervious surface mapping in dense and heterogeneous urban areas.

Share on RLBI navigator
NO

Authors and Publishers

Author(s), editor(s), contributor(s)

Demarchi, Luca
Chan, Jonathan Cheung-Wai
Ma, Jianglin
Canters, Frank

Publisher(s)
Data Provider