Machine vision calibration method for Stewart platform kinematics parameters

The invention discloses a machine vision calibration method for kinematics parameters of a Stewart platform, and the method comprises the steps: placing a camera on an upper platform of the Stewart platform, then moving the upper platform to a specific pose for a plurality of times, shooting an image of a calibration template after each movement, detecting pixel coordinates of cross points on a checkerboard calibration template in an image through an angular point detection method, solving a Perspective-n-Point problem to obtain the pose of a camera relative to the calibration template, finally, constructing an equality constraint kinematics parameter error model, and converting the model into an unconstrained model to carry out optimization solution. According to the kinematics parameter calibration method provided by the invention, the kinematics parameters can be obtained only through the relative pose change under the condition that the transformation relation between the camera coordinate system and the upper platform coordinate system is unknown, and the error of the initial length of the supporting rod can be effectively determined.