Automated UAV-based tool for urban forest tree monitoring using machine learning and image processing

Urban forest management requires accurate and scalable methods for monitoring tree structure and health. Traditional inventory approaches are labor intensive and limited in spatial coverage, while existing remote sensing methods often lack the resolution and adaptability needed for diverse urban environments. This study presents an automated UAV-based framework that integrates machine learning, image processing, and topographical analysis for individual tree detection and characterization. The framework employs YOLOv8 object detection for initial tree identification, followed by slope-based region-growing segmentation that refines canopy boundaries using Digital Elevation Model (DEM) and Digital Terrain Model (DTM) data. A two-stage Otsu thresholding approach addresses small tree exclusion in mixed height canopies. Tree structural attributes including height and canopy girth are derived from DEM/DTM analysis, while vegetation health is assessed using the Modified Green-Red Vegetation Index (MGRVI). The methodology was validated on UAV datasets from Maharashtra and Telangana, achieving 95% detection accuracy in sparse canopy regions and 85% in dense areas. An interactive WebGIS interface enables spatial visualization of tree attributes and health metrics to support data-driven urban forestry decisions. The framework demonstrates adaptability across diverse tree species and canopy densities, offering a scalable solution for automated urban forest monitoring and management.