| The harvesting of fruits and vegetables has been one of the most time-consuming and laborious aspects of the growing industry,research and manufacturing of intelligent agricultural equipment to replace the picking of such labor-intensive work to achieve efficient automated picking of fruits and vegetables,at home and abroad is the trend and research focus.In this paper,the research and application of key technologies such as picking mechanical structure,visual system,action track planning of picking mechanism and integrated control system of picking robot was completed for tomatoes grown indoors in the shed.The details of the study are as follows:(1)The overall design of the tomato picking robot system is presented in this paper.According to the requirements of ripe tomato picking operation,the working principle and basic structure of tomato picking robot is analyzed,and the overall design of tomato picking robot composition:four degrees of freedom articulated robot structure,flexible end-effector,depth visual recognition and positioning system.(2)In this paper,the existing standardized tomato cultivation method is improved based on the working characteristics of the picking robot,and a tandem,multi-articulated,four-degree of freedom robot arm is designed based on the improved working environment.Based on the existing standardized tomato cultivation method,a two-mode automated cultivation method more suitable for robotic picking is proposed by using horticulture method.(3)In this paper,the internal and external parameters of the visual sensor are calibrated and the tomato automatic identification and positioning algorithm is studied.In this paper,a depth visual recognition and positioning system for a tomato picking robot is built on an open source robotic operating system(ROS).To correct the distortion of the depth sensor and ensure the complete projection of the points in the spatial scene onto the image,the calibration of the internal and external parameters of the Kinect V2 camera is completed on the ROS using Zhengyou Zhang’s calibration algorithm,and then to address the problems of complex image processing and positioning errors in the existing algorithm for visual recognition and positioning of multimodal information,a planar image recognition and tracking algorithm is introduced to pre-process the image information,improving the recognition accuracy of the traditional algorithm for visual recognition and positioning of multimodal information,and on this basis,the simulation target is successfully identified.(4)In this paper,a study was conducted on the problem of robotic arm trajectory planning and hand-eye calibration was completed.The determination of the position relationship between the robotic arm and the depth sensor was accomplished by storing the accessible path points in the form of a database using the Rapidly Expanding Random Tree(RRT)algorithm by extending the random tree to eventually form a collision-free trajectory at random,and the ArUco tag on the end of the robotic arm was captured by the Kinect V2 camera on the open source robot operating system(ROS).(5)The paper concludes with the construction of a model machine for a machine vision-based target grabbing test.Using the open source robot operating system(ROS)as the upper control,ArbotiX controller as the bottom control interface of the robotic arm,Turtlebot-Arm as the body of the robotic arm,Kinect V2 camera as the depth vision sensor,to build the picking experimental platform,and carry out recognition and grasping tests,the experimental results show that:machine vision system recognition success rate of 94%,when the recognition distance is less than 600mm,the positioning error is less than 10mm;robot arm positioning accuracy maximum error 2.5mm,average error 1.1mm robot arm track planning success rate of 99%,track execution success rate of 100%;robotic picking individual tomatoes average time 15s,picking success rate is more than 87.7%. |
