Tree detection, delineation, and measurement from LiDAR point clouds using RANSAC.

As Light Detection And Ranging (LiDAR) (point) data sets increase in resolution, earth scientists become more interested in detecting and delineating trees using LiDAR. The majority of conventional methods that detect and delineate trees convert point data into gridded surfaces. Unfortunately, this conversion process has the potential to introduce error. We improve a pointbased geometric model fitting strategy based on “RANdom Sample Consensus” (RANSAC), known as StarSac, and compare the method’s results against field data. The analysis demonstrates that StarSac produces similar results to field data, and is a strong alternative to conventional methods.