Geotechnical experiments represent an interesting class of experiments because they undergo very, very high deformation. This makes them difficult to analyze using standard DIC techniques. For this particular experiment, the data set used was the penetration of a flat footing and an entrapped sand plug into an underlying clay layer as outlined in "Improved Image-Based Deformation Measurement for Geotechnical Applications" by Stanier et al. Dr. Stanier incorporated some of the core routines used in Ncorr into his own software package specially suited for geotechnical experiments which is available on his website.

The nature of this experiment is different from typical solid mechanics experiments because the pattern is generated by using different colored particles. As the flat footing penetrates the clay, the deformation gradients become extreme in the vicinity of the footing. The colored particles tend to move with the displacement of the clay, but tend not to deform (unlike a standard pattern used in DIC for deformation measurement, which typically deforms with the sample). This causes the pattern to change significantly, leading to decorrelation such that the analysis requires many reference image updates..

The resulting Eulerian displacement fields for this particular experiment (400 images) were analyzed using the C++ port and are shown below:

Since the deformation is so high, and so many reference image updates are used, I think this warranted some additional analysis. I decided to apply an image warp in order to update the original reference image and do some comparisons with the final current image. The image warp is applied using the portion of the reference image within the ROI and the Lagrangian displacement fields:

I've taken the liberty of overlapping the highly deformed regions of the warped reference image and the final current image to see the differences, which give a sense of the quality of the overall match, as well as an illustration of the decorrelation effects:

The correspondence is pretty interesting. There's basically a zone under the "flat" that forms a wedge where the deformation is relatively low. The area immediately outside this wedge is where most of the deformation occurs. This is observed as a "stretching" in the warped reference image. But since these particles are rigid, they are simply displaced in the final current image. This is a direct observation of the decorrelation effect. Outside of this region, the deformation is relatively low, and the correlation between the warped reference image and final current image is pretty good.