Research On Crayfish Grasping Based On Machine Vision

Machine vision technology is widely used in various fields,especially in agriculture.Crop image analysis has become one of the trends in the quality measurement of agricultural products,of which machine vision is an important part.Robotic arms have the high accuracy and flexibility to perform tasks of high quality standards.By using vision systems as the eyes of robotic arms,vision algorithms can detect and identify targets,and combine information processing and mechanical control to achieve collaborative work,which can effectively enhance production efficiency and autonomy.Therefore,it is important to expand the application area of robots in order to promote the improvement of productivity and autonomy.First,this paper introduces the basic framework of image processing and neural network theory for crawfish recognition model.YOLOv5 is used to complete crawfish object detection,and the resulting samples are used as a sample library for crawfish body recognition,and a multi-objective tracking algorithm based on YOLOv5 and Deep SORT is explored for machine vision crawfish recognition;the feasibility of the crawfish recognition model is verified through experiments.Next,the basic principles of the image and camera calibration methods are described,including the two-step Tashi method and the Zhengyou-Zhang planar model calibration method.In the experimental session,the ROS-Ubuntu development tool is used to calculate the internal reference matrix,aberration parameters and conversion coefficient matrix of the binocular stereo camera based on the Zhengyou-Zhang planar model method.The kinematics of the robotic arm is then analysed theoretically and modelled based on the D-H method.Using algebraic methods,the front and rear kinematic equations of the robotic arm are solved and an analytical solution to the rear kinematic equations of the robotic arm is obtained.The cubic and quintuple polynomial methods used to plan the trajectory of the robotic arm are analysed and the results are experimentally verified.Finally,experiments with the mechanical manipulator were carried out on the basis of mechanical modelling and the expected results were obtained.We demonstrate the accuracy of the inverse design of the robotic arm and the correctness of the target localisation method,and conclude with some suggestions for further research.