| In this study, an advanced region growing algorithm was developed to automatically delineate individual tree crowns. The data used for this research are the Light Detection And Ranging (LiDAR) data (with a point density of 3.2 returns/m2) and the high spatial multispectral optical imagery (with a spatial resolution of 0.2 m by 0.2 m) collected over a natural forest scene in Ontario, Canada in September 2005. With the advanced region growing method, a template matching technique was first employed to accurately detect the individual tree tops, and they were used as seeds for merging individual segments into tree crowns during the region growing process. In addition, a new strategy was proposed to explicitly take the crown shape into account to control the segments merging. The results show the synergy of the LiDAR and optical imagery can significantly improve the individual tree crown delineation. Compared with manually delineated tree crowns and ground plots, the developed method was able to provide accurate tree crown map for the study and test sites. Specifically, 92% of the tree crown delineations were as accurate as the manual interpretation, and 82% of the tree crowns were successfully delineated. In addition, the developed method was compared with the commonly used commercial software, Definiens. The results show that the advanced region growing method performs better than Definies in terms of integrated tree crowns were generated. |
