Efficient and automated initial value estimation in digital image correlation for large displacement, rotation, and scaling

The initial value estimation for seed point is the first step in digital image correlation calculation. Among the existing algorithms, the Fourier-Mellin transform-based cross correlation (FMT-CC) algorithm is one of the most efficient and robust owing to its rotation- and scale-invariance. However, when the displacement is large (more than a hundred pixels), the FMT-CC algorithm fails. In this paper, an automated and efficient initial value estimation method based on an FMT-CC algorithm is presented to deal with large displacement, large rotation, and large isotropic scaling. The relationship between subset size and the maximal displacement in the FMT-CC algorithm is studied, and a strategy of setting the subset size according to the estimated displacement is proposed to improve the robustness of the FMT-CC algorithm. In addition, in cases of large displacement, a multi-scale search method is proposed to improve efficiency. The experimental results show that the proposed methods can realize rapid and automated initial value estimation even under conditions of large displacement, large rotation, and large isotropic scaling. The computational efficiency of the multi-scale search method is about one order of magnitude higher than the traditional FMT-CC method.