Research On Manipulator Sorting Technology Based On Machine Vision

With the development of made in China 2025 strategy,the application of manipulator in intelligent manufacturing has become more and more extensive.The manipulator which introduces machine vision changes the traditional industrial manufacturing into the intelligent mode,which not only improves the production efficiency,but also changes the conventional production pattern.There are many disadvantages to manually complete the sorting work of different objects.This paper mainly studies the sorting work of manipulator based on machine vision,including the motion analysis of manipulator and visual classification and recognition.The sorting system takes the aubo-i5 manipulator as the main hardware platform,and combines the machine vision to design the software system and carry out the experiment verification of different workpiece sorting,which has high theoretical guidance and application value in the actual production.Firstly,the M-D-H method is used to establish the link coordinate system of AUBO-i5 manipulator and calculate the homogeneous transformation matrix between the corresponding coordinate systems.The kinematic modeling analysis of the manipulator is carried out to obtain the mapping relationship between joint space and Cartesian space,and the trajectory planning of the manipulator with quintic interpolation is completed according to the forward and inverse kinematics.In order to guide the manipulator to sort accurately,the vision system needs to be calibrated.The binocular calibration and binocular vision positioning are completed in Halcon to get the internal and external parameters of the binocular camera.Meanwhile,the eye and hand calibration of Eye-to-hand mode is carried out to get the conversion matrix between the camera coordinate system and the manipulator coordinate system.Secondly,in order to solve the problem of low accuracy of target classification with single type features,this paper extracts the affine invariant moment,circularity and rectangularity of workpiece after image preprocessing,and makes the data set of the classifier by means of multi-feature fusion to improve the accuracy of classification and recognition.In order to solve the problem of incomplete information and low target object recognition in the image collected by the camera in the factory and other environments,the classification and recognition accuracy will be reduced if the quality of the image is too low.In this paper,an improved Retinex method is adopted to enhance the image.Then,aiming at the problem that the classification accuracy of traditional workpiece classification method is low due to the strong nonlinear mapping relationship between the characteristic parameters and types of objects,this paper proposes a method of workpiece classification and recognition based on quantum particle swarm optimization(QPSO)and back propagation neural network(BP).QPSO is used to optimize the weight and threshold parameters in BP network to solve the problem that the traditional particle swarm optimization algorithm is easy to fall into the local optimum.The data set composed of feature vector and category label is trained in QPSO-BP network,and the trained classifier is used to complete the classification and recognition of different objects.Finally,the experimental system was designed with the aubo-i5 manipulator as the main hardware platform,and the QPSO-BP classification method proposed in this paper combined with the image processing software system to complete the sorting of different workpiece by the manipulator.The experimental results show that the method can accurately classify and identify the workpiece in the process of sorting.