Research On Key Technologies Of Robot Carton Palletizing Based On Machine Vision

Nowadays,industrial robots are playing an increasingly important role in assembly line operations.More and more jobs are being done by industrial robots,and more and more jobs are being occupied by them.Meanwhile,how to make robots work more intelligently and efficiently has become a hot topic.Therefore,adding machine vision to the robot and making it independently judge the working environment is an important means to make the robot work more intelligent and efficient.This topic focuses on the robot carton palletizing based on machine vision.According to the implementation steps of visual palletizing,a systematic and in-depth study is carried out.The specific research contents are as follows:The kinematics analysis method of robot is studied.The modified D-H method,which is more universal,was used to model and analyze the kinematics of UR5 robot,and the expression of the forward solution of UR5 robot was obtained.The working space of UR5 robot was drawn by Monte Carlo method.Based on the forward solution,the expression of the inverse kinematics solution is solved by using analytic method,and the simulation is carried out,which lays the foundation for the subsequent trajectory planning.The principle and calibration of binocular vision system are studied.The principle of binocular vision and distortion correction method are analyzed.The projection matrix and distortion parameters of binocular camera are obtained through calibration experiments.The hand-eye calibration of "eye out of hand" is studied.The transformation matrix of hand-eye calibration is obtained by programming calculation according to the image of calibration plate and the calculation method of hand-eye calibration.SIFT algorithm and SURF algorithm which are commonly used in binocular image matching are studied and compared.This paper proposes an improved SURF algorithm with polar constraint and PCA dimension reduction.Compared with traditional SURF algorithm,the improved algorithm reduces the matching time by more than 15%and the accuracy reaches more than 95%while removing most of the mismatching points.It is also verified by image matching experiments.Based on the projection matrix of binocular camera and the matching results of binocular images,the method of 3D reconstruction of spatial points is further analyzed.Two trajectory planning methods for robot motion are analyzed.For joint space trajectory planning process trajectory was studied,and the commonly used because the spline curve has good flexibility and wide application,therefore this paper proposes a cubic B spline curve as optimization object,aimed at the optimal time,by means of genetic algorithm for the trajectory optimization method,the running time of the joints decreased by 14.5%.At the same time,according to the characteristics of the transition of the palletizing trajectory in Cartesian space is not smooth,a palletizing trajectory planning method using arc trajectory smooth transition between straight trajectories is proposed,and the trapezoidal velocity control is used to improve the smoothness of the robot motion.Finally,the communication between the experimental equipment and the computer is established on the ROS platform,and the visual palletizing experiment is carried out,which verifies the feasibility of image processing and trajectory planning,and can be applied to the palletizing link on the production line.