An Improved YOLOv8-based Method for Real-time Detection of Harmful Tea Leaves in Complex Backgrounds

Abstract Tea, a globally cultivated crop renowned for its unique flavor profile and health-promoting properties, ranks among the most favored functional beverages worldwide. However, pests and diseases severely jeopardize the production and quality of tea leaves, leading to significant economic losses.While early and accurate identification coupled with the removal of infected leaves can mitigate widespread infection, manual leaves removal remains time-consuming and expensive. To address this challenge, this paper introduces the YOLO-DBD network model for detecting of harmful tea leaves. The model excels in efficiently identifying harmful tea leaves with various poses in complex backgrounds, providing crucial guidance for the posture and obstacle avoidance of a robotic arm during the pruning process.The improvements proposed in this study encompass the C2f-DCN module, Bi-Level Routing Attention, Dynamic Head, and Focal-CIoU Loss function, enhancing the model's feature extraction, computation allocation, and perception capabilities. Comparative analysis with the YOLOv8s model demonstrates a 6% improvement in mAP and a reduction of 3.3G FLOPs in the YOLO-DBD model.