Comparing digital object based approaches for mangrove tree crown delineation using WorldView-2 satellite imagery

Mangroves are dense, spatially heterogeneous, and have limited height variations between neighbouring trees. Delineating individual tree crowns is thus very challenging. This study evaluated several different object-based image analysis approaches for isolating mangrove tree crowns using a WorldView-2 high spatial resolution image and a digital surface model generated from true colour aerial photographs. The combination of WorldView-2 imagery, a local maxima filtering algorithm, and a region growing approach provided the best accuracy of the methods tested. The inverse watershed method of tree crown delineation showed the lowest accuracy, demonstrating that this method is only suitable for homogeneous forests with reasonable height variations between trees. Therefore, while isolating mangrove crowns using remote sensing is a challenging task, it is possible to achieve good results with very high spatial resolution imagery and appropriately parameterised object-based algorithms.