In the heart of Russia, at the Federal Scientific Agroengineering Center VIM, a groundbreaking study led by Alexey Kutyrev is revolutionizing orchard management. The research, published in the E3S Web of Conferences, translates to the English title “Sustainable Orchard Management Using UAVs: Deep Learning for Apple Detection and Yield Estimation,” is set to transform how we approach agriculture, with potential implications for the energy sector.
Imagine drones soaring over vast orchards, capturing high-resolution images that are then analyzed by sophisticated algorithms to count apples and estimate yields. This is not a futuristic dream but a reality being developed by Kutyrev and his team. By employing the YOLO11 architecture, specifically models from YOLO11n to YOLO11x, they have achieved remarkable accuracy in fruit detection. “The YOLO11x model achieved the highest performance metrics,” Kutyrev explains, “with mAP@50 = 0.816, mAP@50-95 = 0.547, Precision = 0.852, and Recall = 0.766, demonstrating its effectiveness in complex, high-density orchards.”
The process involves creating orthophotos, segmenting data into tiles, and training a convolutional neural network (CNN) with transfer learning and data augmentation. The DJI Mavic 3 Multispectral drone, equipped with a 20 MP RGB camera, captures the images. Data augmentation techniques, including flipping, hue adjustment, blurring, and Tile 8×8 transformation, expanded the dataset from 11 to 2,000 images with 51,797 objects (34,383 apples and 17,414 fallen apples).
The implications of this research extend beyond agriculture. As the world grapples with climate change and the need for sustainable practices, efficient orchard management can reduce waste and optimize resource use. For the energy sector, this means less energy spent on manual labor and more on technological advancements that can drive sustainability. “The method maintains geospatial alignment and visualizes fruit distribution across the orchard,” Kutyrev notes, highlighting the precision and efficiency of the system.
The study also introduces a Tkinter interface that displays detection results and summary data for each orchard section, making it user-friendly for farmers and agronomists. Future work includes integrating multispectral data and 3D modeling to enhance precision further. This research not only automates orchard monitoring and yield assessment but also sets the stage for a future where technology and agriculture are seamlessly integrated, driving sustainability and efficiency across various sectors.
