Deep Learning-Based Estimation of Olive Flower Density from UAV Imagery

Olive cultivation, as a vital industry, faces various challenges, notably yield prediction. The flowering phase of the olive tree (Olea europaea) represents a crucial stage in its life cycle, directly influencing crop yield and fruit quality. This cyclical sequence, from flowering to harvest, is of utmost significance in crafting a decision support system for precise orchard management. This system entails the precise allocation of nutrients at the appropriate time and location, essential for fostering the production of high-quality olive oil. In this study, we propose a deep learning method to estimate the density of pre-anthesis flowers in olive trees using images captured by UAVs. we compare various neural network architectures to evaluate their efficacy in task segmentation. Results indicate that the YOLOv8 architecture excels in segmentation, delivering superior performance compared to other models. Additionally, we also compare methods for estimating the density of pre-anthesis olive inflorescence flowers. The obtained results were compared against experts' assessments (categorized into four levels: upper right, upper left, lower right, and lower left sections), revealing a strong correlation.