| Tomato picking is an important part of the tomato production process.With the growing area of tomato cultivation year by year,the aging of the population and the reduction of agricultural labor force,the development of tomato harvesting robots is of great significance.In the process of tomato picking,the recognition and location of tomato stem and the flexibility of picking manipulators are the key technologies of tomato picking robot,and also the important guarantee for fast,accurate and non-destructive picking.This thesis takes truss tomato picking as the research object,the tomato string-harvesting robot system is constructed,the tomato string recognition and positioning method based on YOLOv3 is studied,and the end effector of a manipulator with integrated clip-shear is developed.On the basis of the above research,experimental research is carried out to verify the reliability of the proposed method.The main research contents of the article as follows:1)A tomato string-harvesting robot system is constructed: According to the characteristics of tomato string harvesting,a tomato bunching robot system is constructed,analyze the composition of each part,and discuss the tomato string harvesting method.2)Research on tomato string recognition and location method: Aiming at the problem that it is difficult to identify and locate the thin stem of tomato string,a method of tomato string recognition and location based on YOLOv3 is proposed.Firstly,YOLOv3 is used for rough segmentation to obtain the target tomato string stem area;K-means clustering method,Open operation and Skeletonization are used for fine segmentation to obtain the image position of the picking point;finally,the large neighborhood mean method is proposed to obtain the picking point depth value,and the picking point image coordinates and depth information are obtained.3)Development of end-effectors: Aiming at the problem of fragile fruit and slender stem of truss tomato,an end-effector with clip and shear integrated was developed;the scissors and the clamping part share the same power through torsion spring to realize the integrated design of clip and shear;the design of inward arc of scissors blade can improve the fault tolerance of picking;the whole body is arranged coaxially and the structure is slender,so as to reduce the interference and damage to branches and leaves during picking,4)Experimental research: an experimental platform built to test the performance of various parts of the tomato string-harvesting robot.In the experiment,the fruit recognition rate is 92.95%,the fruit stem recognition rate is 97.78%,the recognition speed is 98.29 ms/piece,and the hand-eye positioning error was less than 10 mm,the clamping and cutting rate of tomato string is 100%,and the overall picking rate is 90%,which meets the requirements of fast,accurate and non-destructive picking,and improves the picking efficiency. |
