Image Distortion and Correction Caused by Camera Tilt in Machine Vision Measurement

In machine vision measurement, image distortion caused by camera tilt is prone to occur and is often overlooked. This paper clarifies the causes of distortion through mathematical derivation and proposes a step-by-step correction method: first, solve the internal and external parameters of the camera and traditional distortion parameters based on the Zhang Zhengyou calibration method; second, establish an overdetermined system of equations for tilt distortion using the symmetry of dual points of a circular lattice calibration plate, and determine the tilt axis through least squares fitting; further, construct a tilt angle optimization model combined with the principle of perspective projection, and solve the optimal parameters using the binary search method; finally, realize pixel-level coordinate mapping correction based on the tilt axis and tilt angle. This method overcomes the convergence defects caused by multivariable coupling and significantly reduces the algorithm complexity. Experimental results show that the accuracy of four parameters, including the distance from the dual projection point to the principal point, the difference between the upper and lower side lengths of the rectangle, the distance from the image distortion center to the lattice center, and the image distortion degree, has been significantly improved.