MAPPING PERENNIAL FORESTS USING REMOTE SENSING DATA

Abstract

This study highlights a practical and methodological solution for mapping multi-year forest stands based on remote sensing data, specifically drone (UAV) images. The aim of the work is to automatically separate forest contours, row geometry, and individual crown objects in a GIS environment based on high-resolution orthophotos and 3D products (DSM/DTM/CHM and point cloud) obtained by UAV photogrammetry and calculate their inventory indicators. The methodology is based on a chain of flight design and geodetic linking, photogrammetric reconstruction, object-oriented analysis (OBIA), and deep learning (CNN) approaches, as well as a statistical evaluation chain of thematic accuracy (Precision/Recall/F1). The results show that the internal heterogeneity of the forest stand (sparing, gaps, row irregularities) can be reflected in spatial layers and management decisions (replanting, mechanization corridors, resource planning) can be made in an information-based manner.