KSEF RDE: Forest Modeling using Airborne LiDAR Information

Forest studies are currently done mostly manually through field cruises and interpretation of the aerial imagery, which is a labor-intensive task. Airborne lidar technology is capable of building a point cloud from the top surface of the ground objects. Modeling urban areas from the lidar has been successfully accomplished. However, for forest modeling, trees do not much conform to predefined geometric shapes and sizes, and their special distribution is not uniform. That has made the previous related work incapable of providing sufficiently robust approaches. We propose to model forests by developing robust automated procedures for individual tree delineation and attribution of morphological features. These procedures will allow the remote characterization of vegetation over large area, which significantly reduces the cost and manual effort needed, and has major implication in large-scale natural resources issues including remote forest inventories, quantification of biomass and carbon storage, mapping wildfire potentials, etc. Our current emerging work has started with development of a novel tree segmentation algorithm within the lidar point cloud based on the fact that any two distinct objects are very likely to have horizontal gaps in some places between them. So, the basic idea is to try to extend those gaps in order to segment the objects apart. The most trivial advantage of this approach over the previous work is that it does not make any prior assumption about the shape of the objects. Our preliminary result on Robinson Forest data has been very promising. Robinson Forest is a 14,000-ac experimental research forest in eastern Kentucky with high diversity of tree species and complex terrain structure. We intend to continue this research work and publish the novel discoveries in peer-reviewed conferences and scientific journals.