Algorithm for Identifying Tree Crowns from UAV Images Based on a Structural Model

The purpose of the study is to increase the reliability of interpretation of tree crowns from images from unmanned aerial vehicles (UAVs) of forest areas by developing an algorithm using watershed methods with markers and filtering local maxima. During the study, a structural model was developed, built based on an ontology apparatus, including elements of systems of ground objects, their characteristics and relationships, as well as a system of objects in a detailed multispectral image, their properties, and their relationships. The developed software, based on the developed two-pass algorithm, can identify the contours of tree crowns, calculating their characteristics, and recording them in a shapefile. The reliability of the classification of UAV images of forest areas using the developed algorithm is 73.4%, and the reliability of decoding increased by 10.6% compared to standard algorithms.