Accuracy Assessment of the Discrete Classification of Remotely-Sensed Digital Data for Landcover Mapping.
Abstract
Remotely-sensed digital data may potentially help natural resource managers of military installations derive landcover information needed to inventory and monitor the condition of publicly owned lands. One method of deriving landcover information is to perform a discrete classification of remotely-sensed digital data. Before using a remote-sensing derived landcover map in management decisions, however, an accuracy assessment must be performed. This study compared methods of site-specific and non-site-specific accuracy assessment analyses in the context of deriving a general landcover map. Non-site-specific analysis was found to be useful only for detecting gross errors in a classification. Site-specific analysis was found to provide critical information about a classification's locational accuracy. The use of an error matrix was also found to provide additional insight into classification errors, and the use of the Kappa Coefficient of Agreement was found to account for random chance in the accuracy assessment. At a minimum, a Kappa Coefficient of Agreement should he attached to any resultant classification of satellite imagery. Ideally, several measure of accuracy assessment should he performed and included as documentation with any classification. (MM)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1995
- Accession Number
- ADA296212
Entities
People
- Calvin F. Bagley
- Gary M. Senseman
- Scott A. Tweddale
Organizations
- Construction Engineering Research Laboratory