Speaker
Description
Accurate, early and cost-effective field information is a bottleneck for both forest and orchard production systems, especially when decisions depend on reproductive potential and individual tree or hedge structure. The ARGUS consortium develops and validates operational UAV and terrestrial LiDAR workflows to (i) automatically quantify reproductive organs in maritime pine crowns and chestnut orchards, and (ii) estimate dendrometric variables for maritime pine trees and hedge-scale descriptors for hazelnut orchard hedgerows.
For maritime pine flowering, ARGUS implements an end-to-end processing chain from acquisition to indicator delivery. UAV image acquisition was evaluated both with autonomous waypoint missions and targeted crown imaging, and the retained workflow relies on standardized, close-range crown imagery to capture high-resolution images of reproductive organs. A deep-learning detection model, trained with expert annotations and multi-site acquisitions, delivers automated flower counts that are robust, enabling faster, more consistent and scalable estimations. These outputs support monitoring of flowering dynamics and improve decision-making for forest management and seed supply planning. In addition, ARGUS delivers an automated method to detect and quantify chestnut burrs from UAV imagery under variable orchard conditions. The outputs can be produced at orchard and block scales and are directly usable for yield anticipation and operational management.
For dendrometry, terrestrial LiDAR is used to derive structural metrics tailored to each production context. In maritime pine, individual-tree attributes including stem diameter (DBH) and height provide objective phenotyping variables that support genetic selection and evaluation. In hazelnut orchards, hedge-oriented dendrometry (height, width and volume proxies) enables rapid characterization of canopy structure and provides actionable indicators linked to orchard productivity and monitoring.
| Keywords | UAVimaging; TerrestrialLiDAR; Deeplearning; Dendrometry |
|---|
