<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>QuickBird satellite images were processed using object-based analysis to map the spatial structure of seagrass in sandy shoal habitat in the southern Baltic Sea. A three-level ecological model of seagrass landscape, composed of meadows, beds and patches/gaps, was implemented in the multi-scale object domain. Image segmentation was performed at different spatial scales. In order to determine representative scales for bed level and patch/gap level objects, histograms of delineated objects were analyzed. Using object-oriented classification methods, two hierarchically nested maps of seagrass spatial structure were created. The map of patches/gaps was created using the nearest neighbor classification method in the feature space defined by the mean value of band 2 and the value of the proposed seagrass index. Overall map accuracy was 83%. The second map, which depicted the cover density of seagrass beds, was created on the basis of hierarchical relationships between objects at two chosen spatial scale levels. Both maps were exported as vector objects to GIS. Vector-based mapping of seagrass landscape structures at two scales simultaneously provides new possibilities for using landscape metrics and time change detection methods.</p>
</abstract>

QuickBird satellite images were processed using object-based analysis to map the spatial structure of seagrass in sandy shoal habitat in the southern Baltic Sea. A three-level ecological model of seagrass landscape, composed of meadows, beds and patches/gaps, was implemented in the multi-scale object domain. Image segmentation was performed at different spatial scales. In order to determine representative scales for bed level and patch/gap level objects, histograms of delineated objects were analyzed. Using object-oriented classification methods, two hierarchically nested maps of seagrass spatial structure were created. The map of patches/gaps was created using the nearest neighbor classification method in the feature space defined by the mean value of band 2 and the value of the proposed seagrass index. Overall map accuracy was 83%. The second map, which depicted the cover density of seagrass beds, was created on the basis of hierarchical relationships between objects at two chosen spatial scale levels. Both maps were exported as vector objects to GIS. Vector-based mapping of seagrass landscape structures at two scales simultaneously provides new possibilities for using landscape metrics and time change detection methods.

Object-oriented classification of QuickBird data for mapping seagrass spatial structure

Oceanological and Hydrobiological Studies

QuickBird satellite images were processed using object-based analysis to map the spatial structure of seagrass in sandy shoal habitat in the southern Baltic...

Object-oriented classification of QuickBird data for mapping seagrass spatial structure

Publié en ligne: 02 avr. 2009

Pages: 27 - 43

DOI: https://doi.org/10.2478/v10009-009-0013-9

Mots clés
QuickBird, Baltic Sea, GIS, seagrass, seagrass spatial structure

This content is open access.

Object-oriented classification of QuickBird data for mapping seagrass spatial structure

Publié en ligne: 02 avr. 2009

Pages: 27 - 43

DOI: https://doi.org/10.2478/v10009-009-0013-9

Mots clésQuickBird, Baltic Sea, GIS, seagrass, seagrass spatial structure

This content is open access.

Mots clés
QuickBird, Baltic Sea, GIS, seagrass, seagrass spatial structure