Extraction of tree crowns from high resolution imagery over Eucalypt dominant tropical savanna

High spatial resolution satellite imagery provides data that enable the analysis of detailed landscape information, including tree crowns. The inherent characteristics of Eucalypt crowns are challenging to remotely sensed tree crown extraction. This paper develops and applies an object-based tree crown delineation approach suitable for estimating canopy cover of Eucalypts in tropical savannas. A two level segmentation was undertaken upon QuickBird data. Firstly, a broad segmentation masked out non-Eucalypt dominant communities; and secondly, a finer segmentation and ruleset identified seed objects within crowns and then expanded these objects to cover entire crown extents. Of the 1604 tree crowns manually observed within the scene 84.3% were detected by the automated seed identification process. 75% of tree crowns extracted through the region growing process show strong overlap with their corresponding reference crowns. Results indicate the potential of this method for delineating tree crowns from Eucalypt dominant savanna and the use of this information to estimate canopy cover and tree distribution patterns.