Innovative integration of maturity detection and robotic technology for enhanced litchi harvesting

Abstract An integrated approach marries an optimized YOLOv7 deep learning model with a 6-axis robotic arm, featuring a novel end-effector design, to elevate automated litchi harvesting. The model, refined with ACmix attention and Focal-EIoU loss functions, adeptly identifies litchi maturity across diverse orchard settings, overcoming challenges like variable lighting and occlusions. Rigorous field validations and real-world applications demonstrate the system’s efficacy, with an 85.7% success rate in precisely harvesting mature litchis. This fusion of cutting-edge image recognition with strategic mechanical innovation effectively mitigates labor shortages and enhances productivity, marking a significant leap forward in orchard automation.