High-accuracy intrinsic calibration of the line scan camera using rotation-invariant features and cross-ratio invariance

The conventional line scan camera calibration method, which depends on precise displacement equipment and involves high calibration costs, is replaced with a new technique, to our knowledge, based on rotation-invariant point geometry. First, the matrix camera is used to determine the spatial position of the rotation axis, which is fixed in the plane of the calibration pattern. By utilizing the collinear projection characteristics of the calibration pattern rotating around the fixed axis, the rotation-invariant point-also known as the convergence point of the intersection line of the view plane-is determined by combining the geometrical constraints of the feature point with the principle of the invariant intersection ratio. Subsequently, a world coordinate system is established with the rotation-invariant point as the reference. A linear equation system, including parameters such as the focal length and principal point, is constructed according to the projection model of the line scan camera. The intrinsic reference matrix is robustly calculated using singular value decomposition and nonlinear optimization. The experimental results reveal that the method has a calibration repeatability standard deviation coefficient of less than 2% and a reprojection error of less than 0.3 pixels when the rotary axis is fixed, making it more efficient than the traditional motion platform calibration method. This method provides a practical calibration solution for fixed-axis rotary industrial inspection equipment that does not rely on extrinsic motion control devices.