Research On Projective Homography Based Uncalibrated Visual Servoing

Visual Servoing technology can effectively expand the application scope of robots by introducing real-time visual feedback information.It has become an important research direction in the domain of robotics.The uncalibrated visual servoing technology is free of accurate parameter calibration of the robot visual servo system,which improves the robustness of the visual servo system to parameter estimation errors and avoids the cumbersome process of accurate system calibration,thus significantly enhancing the system's adaptability to complex operating environments and tasks.The existing research of the uncalibrated visual servoing is limited to the image based visual servo framework,which has inherent defects in real-time performance and robustness.This thesis proposes an uncalibrated visual servoing method based on the projective homography,which achieves high efficiency and high precision visual servo control as well as trajectory optimization.The main research contents are as follows:A novel uncalibrated visual servo strategy based on projective homography is proposed.By establishing the equivalent expressions of task functions in three-dimensional Euclidean space,image space,Euclidean homography space,and projective homography space,a new strategy of uncalibrated visual servo in projective homography space is proposed,which overcomes the intrinsic deficiencies of image-based uncalibrated visual servo methods in terms of computational efficiency and robustness of task functions.The accurate and robust estimation method of the projective homography are proposed.Through the easy-to-acquire additional images combined with the virtual parallax method,the proportional ambiguity of the projective homography in the calculation process is eliminated.On this basis,an accurate estimation method of radial distortion parameters based on the division model is further proposed,which simultaneously enables the image correction and the accurate estimation of projective homography under radial distortion condition.The trajectory optimization of the projective homography is achieved without system calibration.By analyzing the optimal trajectory of the camera in three-dimensional Euclidean space and the corresponding image trajectory,an analytical expression of the optimal trajectory in projective homography space is proposed.The expression is completely independent of the system parameters,and achieves direct and efficient trajectory planning of the projective homography space,thus optimizes the camera/robot motion trajectory in three-dimensional Euclidean space.An uncalibrated visual servo controller based on projective homography is established.A visual servo task function based on projective homography was proposed.Theoretical analysis proves that the task function has higher accuracy and robustness than the existing uncalibrated visual servo task function;The Jacobian form corresponding to the task function is analyzed.The online estimation of the Jacobian and the corresponding control law are proposed,thus the uncalibrated visual servo control based on the projective homography is realized.The effectiveness of the uncalibrated visual servo method based on projective homography is verified by experiments and simulations.An experiment platform and a simulation platform for uncalibrated visual servo are constructed.The performance of the uncalibrated visual servo method based on the projective homography is analyzed from the three aspects,which are realtime computing efficiency,positioning and tracking accuracy,and the motion trajectory.The results prove the feasibility and superiority of the proposed new strategy for visual servoing.