Automated tree delineation using very high resolution imagery, with a pixel size of less than 1 meter, holds great potential for forest ecology applications. Many of these applications require the identification of individual trees. Delineating individual trees in deciduous forests is a complex task, but groups of shadows distributed on the edge of trees appear to provide a skeleton outline of the trees. Automated route selection using a minimum cost path based on image DN values is a robust approach for connecting these shadow clusters. Similar polygons were identified irrespective of whether the selection of neighbors was based on the minimum total cost of the path, or minimum average cost. The incorporation of texture also resulted in relatively small changes in tree delineation. However, ratioing the original image and its texture was found to be preferable to a more complex incorporation of directional texture measures in the calculation of the path associated with the minimum cost.
展开▼
