Research And Development Of Potato Automatic Pickup Device

Potato is the world’s fourth largest food crop,and Chinese production scale ranks first in the world.In the process of picking,there are problems such as high labor cost,high labor intensity,low level of picking mechanization and poor adaptability of machinery.In order to improve the mechanization level of potato picking in segmented harvesting in Northwest China,this paper has developed an automatic potato picking device and conducted functional verification tests on it,with the following research contents and results:(1)Overall design of automatic potato picking device.Through field research on the potato picking operation environment in Guyuan City,Ningxia,the technical requirements of the automatic potato picking device in terms of structure and function are analyzed,and the overall plan of the automatic potato picking device is designed and determined.The overall scheme of the device is modeled in 3D by Solid Works,and the structural design and parameter optimization of the walking device,potato picking robot arm,end-effector and conveying device are completed to ensure that there is no interference between the components and the potato picking function can be realized.Combined with the picking operation environment and the working principle of automatic potato picking device,the identification and positioning method of monocular camera and photoelectric sensor information fusion is designed,using monocular camera to identify potatoes to obtain the horizontal position information of potatoes,and combined with photoelectric sensor to obtain the depth information of potatoes.(2)Design of walking system and robot arm control system of automatic potato picking device.According to the structure of the four-wheel independent drive of the automatic potato picking device,the differential motion control model of the walking system is established,and the control of the walking device is realized by using PWM wave to drive the wheel motor.The potato picking robot arm based on Cartesian coordinate system is designed,and the forward and reverse kinematic model of the robot arm is established by combining the joint configuration of the robot arm.According to the inverse kinematic model of the robot arm,the motion control of the robot arm is realized;by judging the change of the state information of the photoelectric sensor,the control of the end-effector gripping potatoes is realized,and the design of the robot arm control system is completed.(3)Potato recognition and localization system design based on YOLO v3.YOLO v3 is used to train the potato images acquired for potato recognition,with a Loss value of 0.0819,an average accuracy of 92.58%,a recall of 82.99%,and an F1 value of 0.86.By the Zhang Zhengyou calibration method,the internal parameters of the monocular camera are calibrated to obtain the internal parameter matrix,lateral aberrations and longitudinal aberrations of the monocular camera and the results of the calibration parameters of MATLAB and ROS(Robot Operating System)are compared and analyzed.Based on the ROS system,a hand-eye calibration is performed using the Aruco calibration board using the eye-to-hand approach to solve the transformation matrix.Due to the error of depth information judgment by monocular camera,the depth information is judged with the aid of photoelectric sensor.Combined with coordinate conversion,the design of the positioning system is completed.(4)Potato automatic picking device functional verification test.The function of the automatic potato picking device is verified in the laboratory and field environment respectively.The field environment is more complex than the laboratory,and there is a certain misjudgment in the identification process,resulting in picking by mistake.The success rate of potato identification in the laboratory and field environment is 91.67% and 82.50% respectively,and the success rate of potato picking is 85.00% and 77.50% respectively;during the picking operation,the average time for the robot arm to move to the horizontal designated position is about 4.4 s,the average time for vertical telescoping is about 1.3 s,and the picking action time is 2.0 s,the average time of returning to the placement point is about 4.5 s,and the average time of completing a set of picking action is about 12.2 s.