Research On Agricultural Picking Robot System Based On Visual Recognition

At present,the development of China’s agriculture has made great progress,and has the new characteristics of large-scale production,specialization of planting,diversification of varieties and integration of sales.With the continuous expansion of production scale,how to reduce production costs and improve production efficiency has become an urgent issue in agriculture.Using automated robots for operations is an excellent way to reduce labor costs and achieve efficient production.However,most of the agricultural production in China is performed manually or only with simple tools.The mechanization level is relatively backward,especially in fruit picking.Therefore,the picking cost is relatively high.In this paper,apple picking is taken as an example to study agricultural picking robots,with the support of National Natural Science Foundation of China(31571571)and Key Technology Projects of Industry in Foresight and Commonness(GY2018018)in Zhenjiang city.The main research contents are as follows:Firstly,according to the actual needs of the existing working environment,the crawler-type motion chassis,five-degree-of-freedom manipulator and end effector of the agricultural picking robot are studied respectively.We design the mechanical structure of the crawler platform,the picking manipulator and end effector,and establish the corresponding kinematics model of the agricultural picking robot,thus realizing the mechanical movement required for the picking of the robot.Secondly,according to the characteristics of the mechanical structure of the agricultural picking robot,the overall control scheme of the robot system is studied,the circuit control system and data acquisition system of the crawler platform and end effector are designed,and the appropriate motion control method is selected to control the manipulator.At the same time,the robot control software system is designed,including the motion platform program,the end gripper program,the upper computer communication program,etc.Thirdly,a control method of agricultural picking robot arm based on visual recognition is proposed.At first,the target apple is identified and located by digital image processing technology,and then the obstacle avoidance path planning of picking manipulator is completed based on RRT algorithm and artificial potential field method.At last,an improved fuzzy neural network sliding mode algorithm is used to realize the servo control of the picking robot.Simulation experiments show that this method can effectively improve the response speed of the robotic arm control system,thereby achieving rapid apple picking.Finally,an apple picking robot experimental platform is built,and on this basis,a robot picking apple experiment is carried out.The experimental results prove the effectiveness of the apple picking robot designed in this paper.Compared with the traditional picking robot,the one-time picking success rate of the robot system in the laboratory environment has reached 93.7%,and the working effect reached the ideal level,which could meet the requirements of actual picking work.