thumbnail

Assessing the accuracy of Landsat Thematic Mapper classification using double sampling

International Journal of Remote Sensing

By:
, ,

Links

  • The Publications Warehouse does not have links to digital versions of this publication at this time

Abstract

Double sampling was used to provide a cost efficient estimate of the accuracy of a Landsat Thematic Mapper (TM) classification map of a scene located in the Rocky Moutnain National Park, Colorado. In the first phase, 200 sample points were randomly selected to assess the accuracy between Landsat TM data and aerial photography. The overall accuracy and Kappa statistic were 49.5% and 32.5%, respectively. In the second phase, 25 sample points identified in the first phase were selected using stratified random sampling and located in the field. This information was used to correct for misclassification errors associated with the first phase samples. The overall accuracy and Kappa statistic increased to 59.6% and 45.6%, respectively.Double sampling was used to provide a cost efficient estimate of the accuracy of a Landsat Thematic Mapper (TM) classification map of a scene located in the Rocky Mountain National Park, Colorado. In the first phase, 200 sample points were randomly selected to assess the accuracy between Landsat TM data and aerial photography. The overall accuracy and Kappa statistic were 49.5 per cent and 32.5 per cent, respectively. In the second phase, 25 sample points identified in the first phase were selected using stratified random sampling and located in the field. This information was used to correct for misclassification errors associated with the first phase samples. The overall accuracy and Kappa statistic increased to 59.6 per cent and 45.6 per cent, respectively.

Additional Publication Details

Publication type:
Article
Publication Subtype:
Journal Article
Title:
Assessing the accuracy of Landsat Thematic Mapper classification using double sampling
Series title:
International Journal of Remote Sensing
Volume
19
Issue:
11
Year Published:
1998
Language:
English
Publisher:
Taylor & Francis Ltd
Publisher location:
London, United Kingdom
Larger Work Type:
Article
Larger Work Subtype:
Journal Article
First page:
2049
Last page:
2060
Number of Pages:
12