dicengine
Question about sparse speckle patterns and low gradients in subsets
Hi,
I have a question about the way software determines where to apply subsets to in the region of interest. The tests that I did were with specimens with a dense speckle pattern (pic1), a sparse pattern (pic2) and no pattern (pic 3) respectively as seen in attachments. I am trying to understand how the subset functions, because I was only able to get results from pic1, for the other sequences I could not get results. Why is DICe able to track deformation for the first specimen and not for the others, does it track the displacement between the speckles, or does it track the differences in contrast values per pixel? All mentioned files and their respective solution files are included in the attachments pictures&solution files.zip
DIC, in general (not just in DICe), depends on image gradients created by the speckle pattern. When the speckle pattern is too sparse, there isn't enough image gradient to get an accurate solution. In the case of DICe, there is a setting that checks the SSSIG (a measure of how much image gradients there are in the subset) and only assigns subsets where this threshold is met. If you want to run DICe for images with low gradient levels (the sparse patterns) simply set the SSSIG threshold to 0. I would also recommend using the SIMPLEX method for these cases as the SIMPLEX method does not require a gradient to drive the optimization.
Hi Ra-ux-ops, be also very careful that your specimen is perfectly perpendicular to you camera during the measurement if you are using single camera (2D) DIC. When I look at your picture I'm not sure if that's the case. Non-perpendicular setups can lead to significant errors. There are several papers available on that matter. One I can recommend is the following: "Error estimation in measuring strain fields with DIC on planar sheet metal specimens with a non-perpendicular camera alignment" by Lava et al.
I see, but if we were to speak about accuracy, is it correct to say that it is subpixel accuracy if it is able to determine the gradients correctly and if so how is it quantified?
Hi Ra-ux-ops, the discussion about accuracy and resolution in DIC measurements is something which goes far beyond a post on a github repository. There are a lot of publications on that subject. The camera, setup, lighting, stochastic pattern, DIC algorithm, interpolation method, subset and step size,.... are all affecting your final results. Depening on what you want to do with your measurements, you should or should not have a deep dive into that literature.