Detection and classification of Shiitake mushroom fruiting bodies based on Mamba YOLO

To address the challenges of high labor intensity and low harvesting efficiency in shiitake mushroom production and management, this article presents a novel detection and classification method based on mamba-YOLO. This method adheres to the picking standards and grade specifications for shiitake mushrooms, enabling the automatic detection and quality grading of the mushrooms. Experiments conducted on a self-constructed shiitake mushroom dataset demonstrate that mamba-YOLO achieves precision (P), recall (R), mean average precision calculated at an IoU threshold of 50% (mAP@0.5), and average precision computed over IoU thresholds ranging from 50% to 95% in increments of 5% (mAP@0.5-0.95) of 98.89%, 98.79%, 97.86%, and 89.97%. The classification accuracies for various categories-mushroom stick, plane-surface immature, plane-surface mature, cracked-surface immature, cracked-surface mature, deformed mature, and deformed immature shiitake mushrooms-are 98.1%, 98.3%, 98.2%, 98.8%, 98.5%, 96.2%, and 96.9%. These results indicate that the proposed detection and grading method effectively determines the maturity of shiitake mushrooms and categorizes them based on cap texture characteristics. The network detection speed of 8.3 ms falls within the acceptable range for real-time applications, and the model's parameters are compact at 6.1 M, facilitating easy deployment and scalability. Overall, the lightweight design, precise detection accuracy, and efficient detection speed of mamba-YOLO provide robust technical support for shiitake mushroom harvesting robots.