Design Of Intelligent Nursery Managemen Robot Based On Deep Learning

In recent years,the scale of China’s flower industry has been expanding,which has become one of the important ways for farmers to achieve entrepreneurship,increase income and get rid of poverty.Because potted seedlings have special environmental requirements,they often need to be transported and placed in different areas in the cultivation process,so as to obtain sufficient light and avoid inappropriate weather.In addition,nursery managers often store seedlings in different areas according to their growth status and types,so as to facilitate unified fertilization,watering and other management.Compared with foreign countries,China’s intelligent agricultural equipment technology started late,and the intelligent robot applied in the field of nursery management is still in the laboratory research stage.This paper intends to solve the problem that the current flower industry faces that the seedling management is highly dependent on manpower,and design a multi robot arm collaborative intelligent nursery management robot based on in-depth learning,so as to realize the unmanned and intelligent management of the nursery.The specific research work is as follows:(1)Firstly,starting from the functional task of the intelligent nursery management robot,this paper designs ten cooperative bionic manipulators,draws the structure of each component of the robot with SOLIDWORK software,simulates the grasping and placement process of flowerpot seedlings,and then establishes the motion model of the manipulator to solve the parameter changes of the joint points of the manipulator;Finally,the sequential logic control system of the manipulator is developed to ensure that each manipulator can successfully complete the tasks of grasping,handling and storing the planter seedlings.(2)This paper introduces the tasks and application fields of target detection,analyzes the problems encountered by traditional target detection algorithms,such as low detection speed,poor generalization ability of feature extraction and limited detection accuracy,deeply studies the network structure,algorithm types and detection accuracy of target detection algorithm based on deep learning,and compares and analyzes the performance of each iterative version of Yolo target detection algorithm,which is the most widely used at present.(3)In order to quickly and accurately identify the species and growth state of potted seedlings,this paper develops a potted seedling target detection system based on deep learning,makes a potted seedling detection data set,adopts the latest yolov5 s target detection algorithm in the detection module,and improves the data enhancement,loss function and other modules in the algorithm,It reduces the missed detection rate of long-distance small object targets in dark environment.(4)In order to ensure the running accuracy of the robot,an omni-directional mobile navigation system of intelligent nursery management robot is developed in this paper;The navigation mode based on the complementary information of gyro inertial navigation and visual navigation is adopted.The angle and image information obtained by the robot main control system are transformed into the mobile parameter command of the robot after processing,and sent to the chassis controller through serial port communication.The chassis control system finally converts the received mobile parameter command into the actual rotation parameters of the stepping motor.Through this communication mode,It can ensure the independence and stability of the navigation system to a great extent.In terms of hardware,the combination scheme of stepping motor and omni-directional wheel is adopted,which can make the robot translate and rotate in any direction,freely walk through the narrow environment such as nursery and warehouse,and improve the scene adaptability of the robot.(5)Finally,the field test of the actual operation performance of the intelligent nursery management robot is completed.By analyzing the test results,the advantages and disadvantages in the design process are found out,and the robot is further optimized and improved.The research of this paper will provide effective solutions for the problems of high dependence on labor and low management efficiency in the current nursery management industry,further expand the application scope of agricultural robots,promote the intelligent development of nursery management industry,and provide effective technical guarantee for the actual landing of intelligent nursery management robots.