Research On Positioning Based On Uncalibrated Visual Servo

The robot vision servo system uses a visual image as an input signal to control the robot to complete the servo task.In an image-based visual servo system,an image Jacobian matrix is used to describe the nonlinear relationship between feature changes in the image plane and changes in the angle of the robot joint.Estimating the image Jacobian matrix by conventional methods requires calibration of the kinematic parameters of the robot and the parameters inside and outside the camera,but under actual factory conditions,complex conditions often make it impossible to perform effective calibration.Therefore,the uncalibrated estimation of the image Jacobian matrix has become a hot research direction in the field of visual servoing.Based on the review of uncalibrated visual servo literature at home and abroad,this paper firstly expounds the uncalibrated visual servo theory,and briefly introduces the pinhole imaging model and the robot kinematics model,which are the basis of the visual servo model.Then,the imagebased uncalibrated visual servo algorithm,namely the nonlinear variance least squares method,is analyzed and derived,and the iterative algorithm of the robot joint angle update of the servo controller is obtained.Then,the dynamic Jacobi method is used for online estimation of the image Jacobian matrix in the servo system,and its divergence performance is optimized based on the iterative recursive least squares method.Next,the problem of image processing at the visual input end of the visual servo system is described in detail.Firstly,it analyzes the advantages and disadvantages of various features of the visual system,compares the characteristics and application scope of various features,and decides to adopt the point features.Then use several commonly used filtering methods and grayscale processing to preprocess the image.Next,the edge detection function and the line detection algorithm are used to detect the edge features of the medium object.Finally,after screening and sequential processing,the required point feature coordinates are obtained.Simulations were performed using the Robotics and Machine Vision toolboxes in a matlab environment.The display analyzes in detail the qualitative relationship between the motion of the points in the task space and the motion of the projected point coordinates in the image plane.Then,as a comparison,visual servo simulation experiments were carried out on the visual servo system under the condition that the camera calibration parameters were known.The experiment verified the theoretical correctness of the image Jacobian matrix.Finally,a simulation experiment was carried out based on the uncalibrated visual servo algorithm.The experiments were carried out in two different initial poses,and the experimental results proved the effectiveness of the uncalibrated algorithm.Finally,this paper constructs a visual servo control system for industrial robots using ordinary monocular cameras and computers,designs a visual servo software system,applies an uncalibrated visual servo system to a UR robot.Then,the scene image taken by the camera is analyzed,and the feature matrix of the desired image is established through experiments.The image feedback control module is established by using the feature information between images,and the visual servo controller is designed.Through multiple experiments and program debugging,the construction of a primary robot vision servo system was completed,which made the robot's intelligent capture eliminate the complicated calibration procedure,overcome the uncertain factors of the original vision system,and improve the robot vision system.Robustness expands the range of applications for industrial robots.