Application of five-point cross-ratio invariance in centroid extraction

To solve the problems that both parameter calibration precision and centroid extraction accuracy of feature points are lower in large-scale 3D vision measurement systems, a centroid extraction algorithm based on the restriction of five-point cross-ratio invariance is proposed. Basd on a Gaussian surface polynomial fitting algorithm and regular feature-point array, the five-point cross-ratio invariance on the same spatial straight line is introduced and used as the restriction condition, to extract centroids of feature points. The Levenberg-Marquardt iterative algorithm is utilized to optimize the centroid coordinates and then optimum solution of all points, are obtained lombined with more precise centroid coordinates of feature points, intrinsic and extrinsic camera parameters are calibrated so as to enhance the parameter calibration and centroid extraction precisions in vision measurement systems. The experimental results prove that the proposed centroid extraction algorithm improves both centroid extraction accuracy and stability, compared with the Gaussian surface polynomial fitting algorithm and the 2×2 quadrel algorithm.