Improved spectral approach for continuous displacement measurements from digital images

Abstract

Digital Image Correlation (DIC) algorithms capable of determining continuous displacement fields are receiving growing attention in some areas of research over subset-based DIC methods. Particularly, in mechanical identification applications where high measurement accuracies are sought, the advantage of continuous displacements are appreciated. Within the framework of inverse problems, the unknown continuous displacements may be expressed in terms of linear combinations of basis functions, e.g. B-Splines or finite element shape functions. In this paper, complementary works have been done to make a spectral decomposition of displacement fields functional, which leads to a fast and memory-efficient approach based on Fast Fourier Transform (FFT). The main challenge has been to make the method operational for images and displacements with non-periodic boundaries. The approach has been evaluated on artificial data based on computer-generated images and prescribed displacement fields. Comparisons made between the spectral approach and the one based on B-Splines and nonlinear optimization proves the superiority of the method in terms of reliability and the required computer resources.