Real‐Time Detection and Robotic Picking of Stropharia Rugoso‐Annulata Using Enhanced YOLOv11s

ABSTRACT Stropharia rugoso‐annulata is a highly nutritious mushroom with rising demand, but manual harvesting remains labor‐intensive, inefficient, and costly. This study proposes an automated harvesting method integrating an improved YOLOv11s model with a structured light camera to detect and localize mushrooms. 2D pixel coordinates and depth maps enable accurate 3D localization. To improve the system's performance, we implemented advanced methods during both the hardware calibration and software training phases. For robotic arm calibration, we applied a regularization technique. For algorithm training, we utilized data‐driven learning principles and neural network optimization strategies. These combined efforts served to enhance the final detection and positioning accuracy of the system we built and tested. The enhanced algorithm achieved a mA p @0.50 of 96.7%, mA p @0.50–0.95 of 81.2%, and a detection speed of 6.52 FPS. Robotic picking experiments yielded a 73.9% success rate, demonstrating the system's potential to reduce labor and improve mushroom harvesting efficiency.