For many years, the classification of forest communities and their associated species from airborne data (primarily aerial photography) has largely been through manual interpretation. With the development of digital imagery, options for delineating individual crowns and identifying these to species have become available but automated techniques for grouping these into meaningful community descriptions still requires development. This paper proposes a method that links individual tree crowns (classified to species) using graph data structures and minimum spanning trees. The method was applied to tree crowns delineated and mapped to species using hyperspectral CASI data acquired over Injune, central Queensland, Australia. The resulting clusters were attributed with tree level statistics relating to crown area and diversity and were shown to produce mapping comparable to that of manual interpretation of stereo aerial photography. The study concludes that provided reliable tree level mapping of species can be achieved, automated classification of forest communities can be undertaken.
|Number of pages||12|
|Journal||Remote Sensing of Environment|
|Publication status||Published - 15 Nov 2010|
- Species classification