Details van artikel 135 van 250 gevonden artikelen
Knowledge-based digital mapping of vegetation types in Big Bend National park, Texas
Titel:
Knowledge-based digital mapping of vegetation types in Big Bend National park, Texas
Auteur:
Plumb, Gregory A.
Verschenen in:
Geocarto international
Paginering:
Jaargang 8 (1993) nr. 2 pagina's 29-38
Jaar:
1993-06
Inhoud:
A knowledge-based strategy is utilized to develop a model for performing automated mapping of twenty vegetation cover types occurring within Big Bend National P ark, Texas. Many of the cover types found within this desert region cannot be reliably identified solely on a spectral basis, even on large-scale, aircraft-borne color imagery. Positive identification may be improved, however, by incorporating additional spatial information that may distinguish given cover types on a non-spectral basis. In this study, digital soils and digital terrain data are utilized with spectral imagery from Landsat Thematic Mapper. This knowledge-based strategy is comprised of three primary elements: knowledge acquisition, rules development, and model structuring. Knowledge acquisition identifies the vegetation composition and non-vegetative site characteristics associated with the occurrence of each cover type. Rules development compares and contrasts these characteristics among pairs of cover types and their subsets Model structuring places the presumed digital analogs of these characteristics within a multi-layered classification. After implementing the automated mapping model, its quality was evaluated with an accuracy assessment. Based upon the cover types field-truthed at 142 sites within the park, the model performed at an 72% level of accuracy. For comparative purposes, a traditional supervised (spectral, statistical) classification yielded a 42% accuracy. The superiority of the model is attributed to its incorporation of knowledge-based information; in essence, identification by considering only those cover types likely to occur over given spectral and physiographic conditions.
Uitgever:
Taylor & Francis
Bronbestand:
Elektronische Wetenschappelijke Tijdschriften
Details van artikel 135 van 250 gevonden artikelen
Tijdschrift beschrijving