Research On Recognition-location And Trajectory Planning Of Robot Based On Binocular Vision

At present,with the rapid development of 5G communication technology,artificial intelligence and robotics,the demand of robot flexible manufacturing,intelligent visual sorting,AI interaction and other applications is increasing rapidly.Some scientific research institutions provide specific solutions for these applications.The core technology of robot's practical work is object location,recognition and trajectory planning.From the perspective of market potential and development trend,in the future,vision based robots will be given higher efficiency and recognition accuracy.In this subject,the key technologies of object location,recognition based on vision and trajectory planning of robot are discussed.Firstly,the research status of the key technologies involved in the robot's grasping work at domestic and overseas is summarized.Based on the standard D-H parameters of Kinova robot,the robot kinematics is deduced,the working space of the end effector is calculated.Considering the micro vibration of the experimental environment,the low frequency vibration image of the camera is restored,and the significance of trajectory planning is introduced.Secondly,the imaging and correction principle of stereo vision is introduced.Calibration algorithm is used to solve the camera's internal and external parameters.On this basis,the hand-eye calibration relationship is determined by matrix transformation.The calibration errors of left and right cameras are calibrated respectively to ensure the reliability of calibration results.The disparity map is generated by stereo matching algorithm and the depth map is obtained according to the matrix relation of spatial corresponding points.Then,the algorithm of 3D-reconstruction from RGB-D based on PCL and the technology of clustering and extracting point cloud are discussed.For clusteringsegmentation technology,the algorithm of volume clustering is optimized,which ensures that it converges to the optimal clustering region with the least number of iterations.At the data structure level of point cloud registration technology,the registration algorithm is optimized,the matching process is refined,and the convergence efficiency is improved.Combined with the matching results of the source and target point clouds,the optimal grasping pose matrix of the object is solved.Finally,based on the joint space,the grasping trajectory of the end effector is studied.Its purpose is to ensure that the key points of the end effector are smoothly connected and improve the positioning accuracy.The trajectory tracking controller is established,which makes the actual trajectory approximate the fitting trajectory,and ensures the robust of the trajectory control system.The comprehensive experiment of vision and trajectory planning is carried out to demonstrate the effectiveness of each key technical algorithm of this subject.