Uncalibrated Vision Technology And Its Application And Research In Robot Flexible Detection System

Uncalibrated vision technology is a hot research direction in the field of robotic visual servo control.At present,most of the research on robot visual servo focuses on the accurate calibration of camera calibration model and robot kinematics model,which is achieved by solving the complex Jacobian matrix method.However,due to the limitation of environmental conditions,this kind of accurate calibration is difficult to achieve in practice.Therefore,the research of uncalibrated vision technology without precise calibration system model is of great significance and broad application prospects for the development of industrial production towards intellectualization.Based on the research of robot visual servo and uncalibrated vision technology,this paper describes the structure and classification of uncalibrated vision technology system,deduces the coordinate change,camera model and robot kinematics model of uncalibrated vision system,and extracts features of objects in image by Harris corner detection and Hough transform.In the image-based visual servo control system of robot,in order to maintain the realtime updating of the robot in the process of motion,the complex Jacobian matrix is usually used to represent the highly non-linear mapping relationship between the change of image features of the object and the change of the robot's position and posture.However,the traditional complex Jacobian matrix solving process is complex.In this paper,a genetic optimization method is proposed.In the uncalibrated visual servo control method based on RBF neural network,the visual mapping model between image feature change and robot pose change is studied by using the superior fitting approximation ability of RBF neural network algorithm.The parameters of RBF neural network are optimized by genetic algorithm to improve training efficiency.The simulation model of uncalibrated vision technology for eye-on-hand robot is established,and the simulation experiment of dynamic positioning of robot is completedThe experimental results show that,compared with the traditional inverse Jacobian matrix algorithm,the genetic optimization RBF neural network algorithm in uncalibrated vision technology can eliminate the cumbersome and singular problems in the calculation process of the traditional inverse Jacobian matrix algorithm.At the same time,the proposed algorithm has the advantages of faster convergence speed and smaller error.