Research On Robotic Structured Light Vision System Calibration

Vision sensors can endow a robot with visual awareness and they are one of the most important parts of a whole robot system. A robot collects the two-dimensional iamges of the environment and objects acquired by its vision sensors then analyzes and describes those images to obtain the useful information to recognize the objets and to complete the measurement and control. Owing to the merit of strong anti-jamming capability, high precision, good real-time performance, active control, etc., the linear structured light is especially suitable for applications in the robotic measurement and control task in the complex environment. When a robot employs the getted visual information to complete the measurement and control tasks, the relationship between the vision sensor coordinate system and the robot base coordinate system and the pose of the structured light have to be acquired first, which requires the calibration of the robot hand-eye relationship and the vision sensor. In this dissertation, some issues concerned with the calibration of the linear structured light vision system, the hand-eye calibration and the robotic vision system automatic calibration based on the image visual control were carried out.Firstly, use a set of concentric circles with a pair of orthogonal diameters as a calibration target, a new camera instrisic parameters’ calibration approach based on the orthogonal property of two vanishing points was proposed. After proving the orthogonal property of the two vanishing points producing by the images of the orthogonal diameters in the camera coordinate system, use the Pythagorean Theorem to establish the constraint equations for the camera intrinsic parameters and solved the constraint equations through the linearized method to complete the linear calibration of the camera intrinsic parametrers. The K-means clustering algorithm and least-squares method were employed to compute the vanishing points coordinates, which effectively improved the extraction accuracy and robustness of the vanishing points thus ensuring the accuracy and robustness of the camera intrinsic parameters calibration. Experimental results and error analysis demonstrated that the calibration precision meets the millimeter level measurement requirements.Secondly, the relationship between the linear structured light plane equation coefficients and geometry parameters of the linear structured light vision sensor was given. The measurement range and accuracy of the linear structured light vision sensor was analysed theoretically. The influence of the image feature measurement error and geometry parameters on the structured light vision sensor measurement range and accuracy were discussed, which provided a theoretical basis for the choice of the structural configuration and parameters of a linear structured light. A calibration method based on collinear known distance three-point was proposed to achieve the calibration of the structured light plane in the two-step method.Thirdly, use a set of concentric circles with a pair of orthogonal diameters as a calibration target, two hand-eye calibration methods were presented. One method was based on the camera motion. Orthogonal vanishing points produced by the the target image were used to compute the camera motion. The rotation matrix was expressed by the unit quaternions, and decoupled method was used to compute the roation and translation linearly. The other method utilized camera extrinsic parameters and the target position parameters. Orthogonal geometry property of the two vanishing points and the fixed point and changing pose method were used to calculate them, respectively. Finally, expressing the rotation matrix by the homogeneous transformation matrix, computed the hand-eye calibration model equation to complete the hand-eye calibration. Verification test results showed that the calibration precision can meet the requirements of the millimeter level measurement, and they are feasible and effective calibration methods.Fourthly, the image-based uncalibrated visual control method of the alignment of the linear structured light and a line was described. According to the configuration characteristics of visual system and polar representation of lines, the image Jacobian matrices between the light stripe and the line in the image space and the robot pose in the three-dimension were deduced. The Kalman filter algorithm was used to estimate them during the control process. The equivalent description of the camera intrinsic parameters calibration error was given and the control system stability and control accuracy based on them were analyzed. Based on Lyapunov stability theory, the stability of the continuous and discrete visual control system were analyzed by way of the Lawes Criterion. An automatic calibration method for the linear structured light plane based on image-based visual control was proposed. Under the conditions of the known camera intrinsic parameters, using the line alignment control scheme completed the automatic calibration of the linear structured light plane. The calibration accuracy is better than that of manual method.Fifthly,the image-based uncalibrated visual control method of the alignment of the linear structured light and a fixed point was described. According to the configuration characteristics of visual system and slope intercept parameters of the line, the image Jacobian matrices of the light stripe were deduced and the Kalman filter algorithm was used to estimate the image Jacobian matrices during the control process. An automatic calibration method for hand-eye relationship based on image-based visual control was proposed. Under the conditions of the known linear structured light visual sensor, based on the hand-eye coordiantes transormaiton model, using the line-to-point alignment control scheme completed the automatic calibration of the hand-eye. The calibration accuracy is better than that of manual calibration method.