Automated Date Palm Detection and Geometric Analysis Using UAV Imagery and Deep Learning

الملخص

Accurate inventory and structural assessment of date palm trees are essential for effective agricultural management in semi-arid environments. Conventional field-based surveys are labor-intensive, time-consuming, and often impractical for large plantation areas. This study proposes an automated framework that integrates high-resolution unmanned aerial vehicle (UAV) imagery, deep learning, and geographic information systems (GIS) for the detection, counting, and geometric analysis of date palm trees. A low-altitude UAV survey was conducted over a 3.66-ha date palm farm in Al-Kararim, Misrata, Libya, acquiring 466 aerial images with a ground sampling distance of 1.77 cm. Photogrammetric processing produced high-accuracy orthomosaics and digital elevation models, achieving a total root mean square error of 5.06 cm. A YOLO-based deep learning model was trained on annotated UAV imagery and achieved a mean average precision (mAP@50) of 93.0% with a recall of 95.1%. The model successfully detected 310 out of 360 palm trees (86%), with undetected cases mainly corresponding to very young seedlings with limited canopy development. Geometric parameters, including canopy diameter and tree height, were extracted by integrating detection outputs with elevation data, and unique spatial identifiers were assigned to each palm. All results were compiled into a geospatial database within ArcGIS Pro, enabling advanced spatial and statistical analyses. The proposed framework demonstrates strong potential for scalable and data-driven precision agriculture applications in arid and semi-arid regions.