Research And Experiment On The Method Of Getting The Surface Information Of Citrus Fruit By Inspection Robot

With the continuous development of agricultural intelligence,scholars at home and abroad have developed a number of intelligent harvesting robots for different fruits and vegetables,which are all in the laboratory stage.In order to improve the level of agricultural modernization,accelerate the marketization of the picking robot,and promote the automation and intelligence of the supervision of the citrus fruit growth process,a method of adding patrol inspection module to the picking robot is proposed to improve the use value of the picking robot.This paper mainly studies the intelligent citrus inspection robot based on three-dimensional reconstruction,puts forward the method of obtaining the surface information of citrus fruit by inspection robot and the scoring method,builds the inspection robot software and hardware experiment platform suitable for the operation in the orchard environment,and realizes the machine through three-dimensional modeling of fruit information given by the visual system on the simulation platform The function of human obstacle avoidance movement planning.The robot can complete the inspection of citrus fruit in the orchard environment,accelerate the process of agricultural robot out of the laboratory to a certain extent,and provide a certain reference value for the picking robot to avoid obstacles.The main research contents and conclusions are as follows:(1)The combination of mechanical arm and camera was put forward to inspect citrus fruit and score the surface information of citrus fruit.Based on the research on the inspection methods of inspection robot and the related research on the species and protection of citrus fruit diseases and pests,this paper proposes to obtain the surface information of citrus fruit through the combination of mechanical arm and camera.On this basis,three kinds of inspection schemes for citrus fruit were proposed,namely,the two shooting positions were taken diagonally,the three shooting positions with an interval angle of 120 ° were taken,and the four shooting positions with an interval angle of 90 ° were taken.In order to facilitate the statistical inspection robot's access to the surface information rate of citrus fruit,a region integration system based on the surface information of citrus fruit was proposed according to the disease location.The surface of citrus fruit was divided into 32 parts through 8 longitudes and 3 latitudes,and the results were verified by the hand-held camera.The three kinds of shooting schemes can be used to most of the citrus fruit table Get face information.(2)Build the software and hardware platform of inspection robot.Through the research of fruit on citrus fruit trees,the inspection module is added to the original hardware platform of picking robot,and the hardware platform of inspection robot is built.In the environment of robot operating system(ROS),the simulation platform of patrol robot is built.Through the URDF file of patrol robot model,moveit! The configuration of function package,the analysis of motion space,the communication between robot and inspection camera and the construction of inspection camera interface,and the construction of inspection robot simulation platform.At the same time,the can interface is used to achieve the consistency between simulation and the motion control of the solid manipulator.(3)The obstacle avoidance experiment of the manipulator is carried out on the simulation platform.Firstly,the information of citrus fruit and branches is acquired by Kinect V2 camera,including the calibration of Kinect V2 camera,the calibration of robot hand and eye,and the recognition of citrus fruit and branches by mask r-cnn algorithm.The recognition results are preprocessed,and the information of citrus fruit is acquired by mapping on Kinect camera.The citrus fruit and branch information obtained by Kinect camera are reconstructed on the ROS simulation platform,and the current mainstream robot autonomous planning algorithms are compared.RRT connect algorithm is selected as the robot obstacle avoidance planning algorithm,and the robot joint value of the solved path is interpolated by quintic polynomial to make the robot trajectory smooth,considering the simulation experiment level Due to the slow loading speed of the robot in the rviz interface and the large memory occupied by the robot motion planning,the AABB bounding box algorithm is proposed to simplify the robot appearance model.In the simulation experiment platform,the inspection robot can avoid obstacles to complete the inspection work in the process of motion planning.(4)The experimental results show that the closer the ratio of horizontal to vertical diameter of citrus fruit is to 1,the more complete the outer surface information of citrus can be obtained by photographing.When the inspection scheme of three photographing positions is adopted,the higher the success rate and work efficiency of obstacle avoidance can be achieved while the integrity rate of Citrus surface can be ensured The average movement time of the robot arm is 86.29 s,the success rate of the robot arm in the inspection process is 90.82%,and the success rate of the robot arm in the movement process is 87.29%.According to the conclusion of indoor experiment,the inspection simulation experiment of citrus fruit in orchard environment is carried out.The experimental results show that the inspection scheme of three inspection positions can be used to inspect citrus fruit in orchard environment.