Research And Realization Of Fruit Suction Based On Deep Learning

Fruit picking operation is a key link in agricultural production.Due to the complexity of the fruit picking operation process,it is mainly done manually.Therefore,this research focuses on the intelligent development in the field of picking and focuses on solving several problems in the current picking field.This paper focuses on the automatic picking of apples,focusing on three main problems in automatic fruit picking:fruit identification,fruit location,and fruit picking.For fruit recognition,a deep learning method was used,based on RGB-D image training convolutional neural network model was used for fruit recognition and localization.Using the depth image,the coordinate coordinates of the fruit in the camera coordinate system space are determined by coordinate transformation.In the fruit picking,the coordinates of the target fruit center point in the camera coordinate system are converted into the base coordinates under the ABB robotic arm coordinate system by hand-eye calibration technology,and then the robotic arm is controlled to absorb the target fruit.This study solved three main problems of fruit picking:fruit identification,fruit positioning,fruit aspiration,network recognition progress of up to 81.4%,recognition error of 2mm,and suction error of about 20 mm.The research on apple intelligent picking provides a certain idea and method for the problems currently faced in the field of picking,which is in line with Chinese policy of vigorously developing intelligent industry.