Work

Comparison of Digital Image Processing Techniques for Classifying Arctic Tundra

Year: 2010

Type: article

Abstract: The arctic tundra vegetation classified in the study area, Toolik Lake Field Station, Alaska, was relatively small in stature (with varying species growing in clusters) and must therefore be placed in different communities. This study compared different digital image processing classification techniques, including unsupervised, supervised (using spectral and spatial features), and expert systems. The dataset was a pan-sharpened 5 × 5 meter spatial resolution SPOT image. Accuracy assessments based on field inspections of each final map were performed. The expert system classification yielded the highest overall accuracy of 74.66%, with a Kappa coefficient of agreeement of 0.6725.

Source: GIScience & Remote Sensing

DOI: 10.2747/1548-1603.47.1.78

Authors: Roy S. Stine, Debasish Chaudhuri, Peter Ray, Prasad Pathak, Mary Hall-Brown

Institutions: University of North Carolina at Greensboro, CDI (United States), University of Alaska Fairbanks

Language: English

Topic: Remote-Sensing Image Classification

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: bronze

Funders: National Science Foundation, University of North Carolina at Greensboro

Award: RUI: A Geomorphic-Tropic Hypothesis for Benthic-Pelagic Coupling in Arctic Lakes