Google earth images of date palm orchards in UAE

Large-scale and high-velocity monitoring of perennial plants, such as date palms, is essential for sustainable agriculture and early warning of biotic/abiotic stresses. This paper builds an automated detection pipeline for large-scale date palm tree identifi?cation from open-source satellite imagery, using 0.3-meter resolution Google Earth images over varied UAE sites. We propose a customized deep learning architecture derived from YOLOv11, which was trained on a carefully curated set of 132 images containing over 10,000 annotated objects. Our pipeline utilizes SAS Planet for geo?tagged image procurement, Roboflow for multistage data processing (polygon-based annotation and heavy augmentation using horizontal/vertical flipping, ±23° rota?tion,±15% brightness adjustment, 2.56% noise injection, and stringent validation protocols. The YOLOv11 architecture, which utilizes C3K2 blocks for lightweight feature extraction, SPFF modules for multiscale spatial processing, and C2PSA attention mechanisms, achieved precision of 96% on the test data, surpassing re?cent models (YOLOv8: 93%, YOLO-NAS: 93%). This paper builds an open source, scalable pipeline for palm detection, directly transferable to resource optimization in UAE date palm agriculture while offering a reproducible framework for global arid-region agriculture.