PSI - Issue 5

Francisco Barros et al. / Procedia Structural Integrity 5 (2017) 1260–1266 Francisco Barros et al./ Structural Integrity Procedia 00 (2017) 000 – 000

1264

5

2.3. Frequency based DIC method

There are two stages in the DIC analysis of a stereo system: matching a subset from an image from a camera to its corresponding point in the other camera (i.e. stereo matching), and finding corresponding subsets in images from different positions or stages of deformation taken by the same camera. The approach used was to perform the stereo matching on the images of the reference state, and subsequently, for each camera, compute the displacement of the subsets for the images taken by each camera at different times. The positions of the 3D surface points at each deformation state were then calculated from their position in both cameras and the camera calibration parameters taken from VIC-3D, and the displacement field was obtained by the subtraction of their coordinates. For stereo matching, rectified images were used, since it was found that this reduced the number of subsets where a reliable correlation could not be calculated. It has the additional advantage of narrowing down the region of the image to search, as rectification guarantees that a point will theoretically be in the same row in both images. Before considering image subsets, correlation was performed on the whole image to find the rigid body displacement of the region of interest from an image to the other. Later, when working with subsets, the correlation is calculated between a certain subset on the first image and a subset on the second image whose position in the image are given by the sum of the first subset’s coordinates and the rigid displacement. The cross-correlation peaks were obtained using discrete Fourier transform (DFT) image registration, and correspond to the peak of the inverse DFT of the spectrum calculated through (6). They were computed with subpixel accuracy, using the single-step DFT approach proposed by Guizar-Sicairos et al. [7], and considering only a rigid translation of the subset. For stereo matching, a result from phase correlation was provided as an initial estimate, as this was observed to make correct identification of the correlation peak more frequent.

3. Results

The results were obtained using a subset size of 29 pixels, with a step of 10 pixels, both horizontally and vertically. The region of interest chosen is shown in Fig. 3, on the reference image from the left camera.

Fig. 3. Region of interest (limits in red), shown on the reference image from the left camera.