Abstract:
Digital Image Correlation is now a widely used method for measuring deforma tion fields with high spatial resolution. Its single camera (two-dimensional) application is used to evaluate in-plane deformation of a planar surface. A subset based 2D DIC al gorithm divides the region of interest of both reference and deformed image into equally spaced subsets and compares them. Matching each subset over the undeformed and deformed images yields the displacement field over the subset grid. Finally, differen tiation of the displacement field yields local deformation and rotation parameters. In conventional DIC applications every subset in the grid is evaluated with the same algo rithm regarding interpolation and optimization functions. This results in a statistical loss of data points in the DIC grid, in correlation due to factors like weak pattern, poor choice of initial conditions and other operational parameters. This loss might be in the form of matching failure (no data), and wrong evaluations (outlier data). The major aim of this project is to develop adaptive DIC algorithms that can recover the most out of the data and minimize the number of lost data points through recursive analysis. These algorithms will detect outliers and reevaluate faulty points using different ini tial guesses and parameters until a reasonable measurement is possible at each subset. This more advanced application proposes to recover DIC points that will simply not work with the traditional methods. However, especially in microscopic length scales certain subsets are not physically analyzable regardless of the algorithm, since their DIC pattern is significantly altered or include little to no distinguishable information. This thesis further aims to use a-priori pattern quality measures to detect and factor out weak patterns and pattern alterations in each subset.
