基于机器视觉的设施葡萄自动采收机设计与试验OA

To address the low level of mechanization level in the harvesting of protected vineyards and the insufficient operational efficiency of existing grape-harvesting robots,this study developed an automatic harvester for protected grapes based on machine vision.An average cutting-position localization algorithm utilizing cluster positional information,along with a corresponding control strategy for the harvesting execution system,was proposed.A grape cluster detection model was established using YOLOv5.Field experiments were conducted to determine the optimal operating parameters and to evaluate the harvesting performance systematically.The results showed that the detection model achieved an F1-score of 0.95 and a mean average precision(mAP)of 0.98,with a field recognition accuracy(Rr)of 95.35％.The harvester could position its cutting system while the chassis was moving forward,enabling continuous harvesting.The optimal performance was obtained at a forward speed of 1 km/h,yielding a harvesting success rate(Rh)of 91.68％,a damage rate(Rb)of 3.99％,and a picking efficiency(Re)of 0.91 s per cluster.This study provides an innovative solution for intelligent,efficient,and low-damage harvesting of protected grapes,offering a valuable reference for advancing mechanization in this field.