Uncalibrated Visual Servo System Based on Kalman Filter Optimized by Improved STOA

The traditional visual servoing of robot is difficult to calibrate the parameters accurately. At present, Kalman filter is used to estimate the image Jacobian matrix, but the noise parameter setting is not appropriate, so the servoing is not accurate. The traditional STOA(Sooty Tern Optimization Algorithm) algorithm is improved by using the adaptive search radius strategy to improve its local convergence ability, and the improved STOA algorithm is used to optimize the noise parameters of Kalman filter, and the penalty term coefficient of the objective function is improved to find the optimal noise parameters under the current system model. Based on the optimal noise parameters obtained in this paper, the noise parameters are applied to the robot eye-in-hand structure, and the image Jacobian matrix is estimated online at each moment, so as to realize the calibration free visual servoing of the robot.