Research on Precise Apple Harvesting Algorithm Based on YOLOv5s

Accurate identification, localization, and ripeness assessment of apples in orchards are essential for yield and quality optimization. An enhanced YOLOv5s target detection model is proposed for precise apple picking management. Through the incorporation of channel and spatial attention mechanisms, the model's feature extraction capability is optimized, thereby enhancing detection accuracy in complex backgrounds. Experiments conducted with apple image datasets featuring leaf occlusion, branch occlusion, and fruit overlap demonstrated that the improved YOLOv5s model effectively detects, localizes, and assesses apple ripeness. Results indicate high accuracy and efficiency, particularly in complex or ambiguous scenarios. This investigation supports agricultural robotic picking tasks and corroborates the efficacy of the attention mechanism in target detection.