| The environment for tomato picking is poor in comfort and has high work intensity,while improper robot picking methods can cause adverse effects such as damage to the fruit skin and reduce the quality of harvested fruits.In response to this issue,a robot picking tomato was taken as the research object,and key technologies such as tomato detection,recognition,spatial positioning,end effector design,robotic arm application,and system integration were systematically studied.A robot that can automatically identify,locate,and pick tomatoes was developed,with intact tomato stems and undamaged skins.The main work and research content are as follows:(1)Collect tomato dataset images with different light intensities.A survey was conducted in the tomato greenhouse of Wuhan Academy of Agricultural Sciences in Hubei Province,and experimental fields were cultivated to plant tomatoes.Images of tomato fruits were collected using binocular depth cameras,and the changes in tomato images under different natural and supplementary lighting intensities were analyzed.The dataset was then enhanced and labeled using the labeling tool Labelme.The dataset was constructed based on different lighting intensities.(2)Research on tomato target detection and recognition and binocular vision positioning technology.The YOLO algorithm was used to train images under different lighting intensities,and a visual system based on YOLO v3 tiny detection and recognition of tomatoes was determined to solve the problem of misidentification caused by fruit occlusion at different maturity levels in fruit clusters.Based on the camera calibration internal and external parameters,the conversion relationship between image coordinate system and spatial coordinate system is obtained,and the depth data corresponding to image pixels is extracted.Conduct localization feasibility analysis experiments to determine the appropriate distance for detection and recognition.(3)A contactless end effector was designed and the motion trajectory of the robotic arm was simulated.Determine the single fruit picking method based on the growth characteristics of tomatoes.Analyze the force on the fruit stem to avoid slipping during shearing,increase the main stem slot to solve the problem of tomato jamming during picking short fruit stems,and reduce the load through 3D printing manufacturing.Use the Arduino control board to integrate communication functions and conduct communication control testing.Modeling and simulation of the robotic arm D-H using MATLAB,using Cartesian space trajectory planning to reduce the scanning coverage area during movement and prevent skin injury during harvesting.(4)Trial research on a prototype tomato picking robot and conduct validation tests.The tomato picking robot consists of five parts: a vision system,a robotic arm,an Arduino control board,an end effector,and a mobile platform.The software integration is based on the ROS robot system.Validate the proposed harvesting technology for autonomous planning of robotic arm movements based on a visual system.The tomato picking robot vision system in this study has a good recognition effect under the illumination intensity of 20000-30000 Lux,with an accuracy of 90.5%,a false recognition rate of 0%,a missed recognition rate of 9.5%,and an average image recognition time of about 1.0s/frame.Tomato picking robots can obtain tomatoes that retain their pedicles and have undamaged skins.The average picking time for a single fruit is about 9.5 seconds,and the success rate of picking is 83.3%. |
PDF Full Text Request