Research On The Robotic Arm Sorting System Based On Machine Vision

Robots play an important role in industry,daily life and the process of artificial intelligence,increasing people are freed from tired and tedious work.The introduction of vision makes robots more intelligentized,and to promote them make timely adjustments to fit in with the changes of external environment.So robots can be applied to more scenes.The robotic arm based on machine vision has become one of the hot research fields.This thesis studies the technology of machine vision and robot involved in robotic arm sorting system,and builds a sorting system using Mitsubishi RV-13 F six-axis robotic arm,then uses components as experimental objects to verify the feasibility of the method.In order to establish the connection between the image coordinate and the robot base coordinate,firstly,the mathematical model of Mitsubishi RV-13 F robotic arm is built.Meanwhile,the kinematics of the robotic arm is analysed,and the solution method of each joint angle is deduced in detail,the selection rules of the optimal solution also are given.Secondly,Zhang Zhengyou's calibration method is used to calibrate the camera and solve the camera's internal parameters,external parameters and distortion coefficients.Finally,the Eye-in-Hand system is calibrated,and the relative positional relationship between the camera and the robot is determined.Because the acquired image will be affected by the scene light and the reflection of the components,the three-dimensional Otsu image segmentation algorithm which with good anti-noise performance is selected to segment the image of components.However,the three-dimensional Otsu algorithm has high time complexity and long operation time,this thesis firstly reduces the gray level to narrow the size of the three-dimensional histogram,secondly,uses the one-dimensional Otsu algorithm to set the upper and lower limits and reduce the space of solution,finally,uses the Cuckoo Search algorithm to optimize the space of solution.The experimental results show that the method can greatly reduce the running time of the algorithm.Meanwhile,a method of dividing pixels into noise and non-noise is provided by thinking about the problem of the wrong segmentation of three-dimensional Otsu algorithm.The experimental results show that the distribution method improves the segmentation effect of the algorithm.Finally,the improved three-dimensional Otsu algorithm is used to segment the image of components,and components can be well segmented.Aiming at the problem that the representation of local features is not prominent enough by CNN.A components recognition method which fuses LBP features and CNN features is proposed,the method firstly uses CNN to extract the deep features of the components,then uses LBP to extract the local features of the components.Finally,the two features are merged in the fully connected layer of CNN.The experimental results show that the method can improve the accuracy of recognizing components.In order to improve the sorting efficiency of the robotic arm to the components,a time optimal trajectory planning method based on the 3-5-3 polynomial is proposed.The method regards velocity as a constraint and the sum of the 3-segment polynomial's time as the objective function,and uses the Cuckoo Search algorithm to solve the optimal time.The experimental results show that the method can solve the better sorting time and improve the sorting efficiency of the robot arm.Finally,the robotic arm sorting system based on machine vision is designed and implemented,and the upper machine interface is designed by QT.The feasibility of the system is verified by the experiments.