Research On Tea Shoot Picking Device Based On Parallel Manipulator

With the development of economy and society,there is an increasing demand for high-end tea made from the young shoots of tea leaves such as sole buds and one bud and one leaf in the market.At this stage,the tea picking machines in the market are not able to pick the tender buds selectively and efficiently,while the efficiency of manual picking cannot keep up with the development speed of the industry,which seriously restricts the industrialization and scale development of high-end tea.With the development of technologies such as image processing,deep learning and parallel robotics,new ideas are provided for the refinement and mechanization of tea shoot picking.In this project,we design a new type of automatic tea bud picking device to identify and pick tea buds by collecting images in the working area and getting the target location through a real-time tea bud recognition model.And then the control system controls the parallel robot to place the picking end to the designated position to pick the tea leaves and store them in the tea collection box.The specific research contents are as follows:(1)Measure the width and height of the tea monopoly in the tea garden and other data,and use them as dimensions to design a parallel manipulator based on linear guide mechanism.By controlling different rotations of the three sets of motors,the slider mounted on the linear guide rail moves different distances to drive the movement of the three sets of follower arms connected to the slider,thus realizing the movement of the mechanical manual platform.After designing the 3D model of the manipulator,the workspace analysis of the manipulator was carried out to verify the rationality of the device design.For the picking end mounted on the mechanical manual platform,firstly,the dimensions of different types of tea leaves of single bud,one bud and one leaf and one bud and two leaves were measured to obtain some design dimensions of the picking device,and then suitable picking ends were designed and processed to be mounted on the moving platform of the parallel manipulator for picking the target tea leaves.(2)A tea shoot recognition model was established based on deep learning.Tea leaf images were captured using a camera to build a dataset for the learning training of the network.After the dataset was built,the Yolov7 network and the lightweight improved Yolov5 network were selected as the neural networks for building the tea shoot recognition model.By comparing the recognition effects of the two networks and combining with the later target localization,the improved Yolov5 network was selected to build the tea shoot recognition model.(3)This paper uses binocular ranging to obtain the 3D coordinates of the target in the world coordinate system.The parameters of the left and right eyes are obtained by calibrating the binocular cameras,and then the parallax of the left and right eyes is calculated by using the SGBM semi-global matching algorithm,and then the depth information of the image and the 3D coordinate values in the coordinate system of the left eye camera are obtained.At the same time,in order to get the coordinate values that can be input into the robotic control system,the hand-eye calibration method of eye-in-hand is used to get the 3D coordinate values located in the robotic arm coordinate system.(4)In this paper,the STM32F103RCT6 microcontroller is used as the control core to write the control program of the manipulator.The motion of the robot arm is controlled by controlling the number of revolutions of each motor on the three active guide mechanisms of the robot arm,so that the picking mechanism moves to the specified position.Meanwhile,in order to correct the coordinate solution part of the control system,the coordinate solution accuracy of the existing control system was analyzed by marking the coordinates of different positions of the mechanical manual platform and comparing them with the coordinates input into the control system of the manipulator.The results show that the motion error of the robot is basically within 3 mm,which meets the requirements of use.(5)By observing the distribution of tea leaves in the tea garden in the field,the path planning problem in the picking process was studied.The path optimization problem of tea picking was transformed into the TSP traveler problem,and the improved ant colony algorithm model was established based on the ant colony algorithm which has a better solution to the traveler problem,and improvements were made in the picking strategy,parameters of the algorithm and the pheromone update method.The algorithm was used to perform the optimal path finding for tea shoot picking under different distribution number distributions.By analyzing the experimental results,it was found that the improved ant colony algorithm was about 2% lower in the distance of the optimal path and about 31.8% lower in the number of iterations than the original version of the algorithm.It can be obtained that the improved ant colony algorithm can get better optimal path results.(6)By combining software and hardware,a tea shoot picking test device was built and the practicality of the device for picking was analyzed.According to the test results,the device can identify 84.4% of the target tea leaves and pick 96% of them successfully.The performance of the device was stable during the picking process and basically achieved the expected design effect.