Drones are now widely used for high-throughput phenotyping in agriculture and forestry.
A forest area of 10-12 hectares with 8,000 to 10,000 trees can be surveyed at cm resolution within 30-40 minutes by a single drone and capture structural and leaf spectral information. These data can be used to estimate growth, for species classification, for detecting vegetation stress and tree mortality, and for assessing forest health. A bottleneck for a wider and routine application of drone imagery is the ability to identify and delineate crowns of individual tree. This task is often accomplished in a tedious manual process or with algorithms that lack precision and accuracy. Here we present a novel fully automated algorithm that accurately delineates individual spruce crowns even from a complex forest canopy. This approach facilitates the estimation of tree-level physiological and phenological parameters. The tree crown delineation tool is a critically important for enabling the first steps in a novel fully automated remote sensing image processing pipeline that is currently under development.
