Introduction: Due to the small size of citrus stems, their color similarity to the background, and their variable position relative to the fruit, accurately locating picking points using robots in natural environments presents significant challenges.Methods: To address this issue, this study proposes a method for segmenting citrus fruits and stems based on an improved YOLOv8n-seg model, combined with geometric constraints for stem matching to achieve accurate localization of picking points. First, all standard convolutions in the model are replaced with GhostConv to reduce the number of model parameters. Furthermore, a convolutional block attention module (CBAM) and a small-object detection layer are introduced to enhance the model’s feature representation and segmentation accuracy for small objects. Then, by incorporating the positional relationship between the fruit and the stem, constraints are defined to match the target stem, and an algorithm is designed to determine the optimal picking point.Results: Experimental results show that the improved YOLOv8n-seg model achieves recall rates of 90.91% for fruits and stems, a mean average precision (mAP50) of 94.43%, and an F1-score of 93.51%. The precision rates for fruit and stem segmentation are 96.04% and 97.12%, respectively. The average detection rate of picking points reaches 88.38%, with an average localization time of 373.25 milliseconds under GPU support, demonstrating high real-time performance. Compared with other models, the improved YOLOv8n-seg model shows significantly better performance.Discussion: This study confirms the reliability and effectiveness of the proposed citrus picking point localization method and lays a technical foundation for the automated harvesting of citrus fruits.

Image moments-based visual servoing control of bagged agricultural materials handling robot

Key technologies of tomato-picking robots based on machine vision

Introduction: Fruit harvesting in natural orchards remains challenging because target fruits are distributed in cluttered and unstructured environments. In the pre-grasp approach stage of multi-fruit citrus harvesting, three issues are particularly critical: limited demonstration data, target ambiguity, and the need for stable and precise local approach motions. Methods: To address these issues, this study proposes a Target-Conditioned Flow-Matching Policy (TCFM Policy), which integrates image observations, robot state history, and explicit target geometric conditions, uses a dual-branch visual representation to encode both global scene context and local end-effector details, and predicts future multi-step TCP trajectories through conditional flow matching. To reduce overfitting to global appearance under small-sample conditions, a target-oriented visual augmentation strategy is further introduced for the global branch during training. Results: A real-world dataset containing 160 valid demonstration episodes was collected on a UR5-based citrus harvesting platform using VR teleoperation. In 50 target-specified multi-fruit trials, the full model achieved a success rate of 76%, a target-picking error rate of 4%, and a picking-point offset rate of 20%. Discussion: A fairness-aligned comparison with a target-conditioned diffusion-policy baseline further shows that the proposed method achieves lower offline trajectory error and better online target-specified approach performance under the same training setting. Ablation results indicate that the ROI branch mainly improves final alignment, while the target-oriented augmentation mainly improves target consistency. These results indicate that explicit target conditioning, dual-branch visual encoding, and conditional flow matching jointly support accurate target selection and relatively stable pre-grasp approach execution in small-sample multi-fruit citrus scenes.