Geographic Object Based Image Analysis (geobia) for Mangrove Tree Crown Delineation Using Worldview-2 Image Data

Providing spatially explicit information of mangrove tree crown is essential to support management and monitoring of mangrove environment. This study developed a method to delineate individual Avicennia marina mangrove tree crown in Whyte Island (Brisbane, Australia) using the newly emerged GEOBIA approach and pan-sharpened WorldView-2 (WV-2, 0.5m pixel size) data. We used region growing technique modified from Culvenor (2002) to isolate the tree crowns based on the tree crown radiometric topography concept. A rule set was specifically developed to (1) enhance the differentiation between trees and canopy gaps, and (2) find the tree crown seed (i.e. tree top), (3) grow the seed towards the tree crown border within the tree class, and (4) refine the tree delineation. The results show that the principal component (PC) 1 and 2 of the image enhanced the differentiation of trees and canopy gaps. The near-infrared1 (NIR1) band of WV-2 found to be best suited for identifying tree seeds and growing the tree boundaries. We compared the GEOBIA delineation result with the manual tree crown delineation based on a very high-spatial resolution aerial photograph (0.75m pixel size) and found a realistic tree crown polygon boundaries with an overall accuracy of 68%. Some delineation errors were attributed to a single tree with multi-canopy crowns (i.e. it was identified as more than a tree) and small mangrove shrubs which were unable to be delineated. The results demonstrate the potential of GEOBIA approach to delineate mangrove tree crowns, but need some improvements in the future. The use of high-spatial resolution image data with pixels significantly smaller than the tree canopy size is an essential requirement for tree crown delineation.