Extraction of Olive Crown Based on UAV Visible Images and the U2-Net Deep Learning Model

Olive trees, which are planted widely in China, economically significant. Timely and accurate acquisition of olive tree crown information is vital monitoring growth accurately predicting its fruit yield. The advent unmanned aerial vehicles (UAVs) deep learning (DL) provides an opportunity for rapid parameters the crown. In this study, we propose a method automatically extracting (crown number area tree), combining visible-light images captured by consumer UAV new model, U2-Net, with deeply nested structure. Firstly, data set (OTC) was constructed, further processed ESRGAN model to enhance image resolution augmented (geometric transformation spectral transformation) enlarge increase generalization ability model. Secondly, four typical subareas (A–D) study were selected evaluate performance U2-Net extraction different scenarios, compared three current mainstream models (i.e., HRNet, U-Net, DeepLabv3+) remote sensing segmentation effect. results showed that achieved high accuracy numbers mean intersection over union (IoU), overall (OA), F1-Score 92.27%, 95.19%, 95.95%, respectively. Compared other models, IoU, OA, increased 14.03–23.97 percentage points, 7.57–12.85 8.15–14.78 addition, had consistency between predicted measured crown, it lower error rate root squared (RMSE) 4.78, magnitude relative (MRE) 14.27%, coefficient determination (R2) higher than 0.93 all subareas, suggesting extracted best profile integrity most consistent actual situation. This indicates UVA RGB can provide highly robust result crowns helpful dynamic management orchard trees.