Design And Research Of Tomato Picking Robot Based On Fruit Recognition

In the process of tomato picking,Accurate recognition of tomato fruit and accurate positioning of tomato stem are the premise and guarantee for the normal and accurate operation of tomato picking robot,and also the key technology in the process of research.Taking tomato picking as the research object,this paper puts forward a set of design schemes for tomato picking robot.The forward and inverse kinematics,workspace and trajectory planning of the picking manipulator were studied and analyzed.For fruit recognition,a recognition method based on the combination of YOLOv4 recognition algorithm and traditional image recognition algorithm was proposed,and the hardware and electrical control system of the robot were designed.On this basis,the reliability of the design was verified by the experimental analysis of the simulation software and the real object.The research contents are as follows:(1)Aiming at the target fruit and actual function of the picking robot,the overall design of the picking robot was carried out,and the selection of its mobile platform,mechanical arm and vision scheme was carried out.(2)The D-H method was used to model the manipulator,and the forward and inverse kinematics of the manipulator were analyzed.The inverse kinematics of the manipulator is obtained using the inverse transformation method.In order to ensure that the manipulator can reach the working position,the Monte Carlo space method is used to analyze the working space of the manipulator,and it was verified by simulation in matlab.In order to ensure the smooth operation of the manipulator,the trajectory planning was carried out.(3)The methods of fruit recognition were studied and analyzed,and a fruit recognition method based on the combination of depth learning YOLOv4 algorithm and traditional image recognition algorithm is proposed.First,YOLOv4 was used for coarse segmentation of the collected image to obtain the fruit target position and the fruit stem position,and then K-means clustering,open operation and skeletonization were used for fine segmentation on the image to accurately locate the position of the picking point.(4)The hardware and control system of tomato picking were designed.The hardware and circuit of the system are determined,mainly including the main controller with STM32F103VET6 as the core and the main control microcomputer with raspberry pie 4B as the core.(5)For the designed and assembled experimental prototype,the dynamics simulation software was modeled and the dynamics experiment analysis was carried out to ensure the stability of the design and driving.Then the experiment of physical picking was carried out.The actual results show that the robot designed and assembled can carry out the tomato picking operation stably and accurately,which verifies the rationality of the whole design scheme.