Research On Control System Of Apple Picking Robot Based On Machine Vision

China is the largest apple producing country in the world,but apple picking mainly relies on manual labor at present,which is time-consuming and increasingly expensive.So it is urgent to realize the mechanization and automation of orchard picking operations.With the support of National Natural Science Foundation of China(31571571)and Zhenjiang Modern Agriculture Project(NY2015025),this project studied the control system of a new generation of apple picking robot based on machine vision,which further promoted the practicability and industrialization process of the picking robot.The main content includes the overall system structure research and mathematical modeling overall research of hardware and software system of picking robot,the apple target recognition and positioning in the complex background,the research on the efficient servo control method of the picking manipulator,and the experimental research on the picking robot.The specific work is as follows:First,in order to adapt to the complex orchard working environment,the basic structure of the crawler type moving chassis,the five degree of freedom mechanical arm and the end actuator of the apple picking robot are studied.According to the actual needs of the existing working environment,we design the electric control system and the data acquisition system of the crawler platform and the end actuator,establish the kinematics and dynamics equations of the picking manipulator.The simulation experiment is carried out.Secondly,according to the basic structure and hardware system characteristics of apple picking robot,the overall control scheme of the robot system is studied.Then,the distributed system research method is used to design the control software of crawler type moving chassis,the control software of end actuator gripper and cutter,the upper computer and the communication system software respectively.At the same time,visual positioning and motion planning algorithm are used to accelerate Response speed of the overall data of the system.Thirdly,the object recognition and location method of apple picking robot based on deep learning is improved.After analyzing and improving the existing network structure of yorov3,we combine the deep learning method with the network structure of yorov3,and adopt the experimental method to enhance the adaptability of apple picking robot to the complex environment,so as to effectively improve the speed of Apple target recognition and location,the effective target recognition rate,increase the picking efficiency.Fourthly,the project improves the servo control method of apple picking robot based on fuzzy neural network sliding mode control.After analyzing the kinematic and dynamic relationship between the joint angles of the manipulator of apple picking robot,the improved fuzzy neural network sliding mode control algorithm is used to realize the servo control of the robot.The simulation experiment shows that the algorithm can effectively improve the dynamic performance of the manipulator control system in the process of apple picking.At last,we complete the experiment of apple picking robot system,because the experiment platform of apple picking robot is built in the simulated orchard environment.The experimental results verify the effectiveness of the proposed method.Compared with the previous picking robots,the success rate of the picking robot system designed in this paper has reached 92.9%,and its working effect has basically reached the ideal level,which promotes the practicability and industrialization of the domestic apple picking robot.