Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
NCTM or The Math Forum.



Re: Crack detection
Posted:
Dec 12, 2012 4:45 PM


"Mee2XuH " <Mee2XuH@gmail.com> wrote in message <kaade8$gt4$1@newscl01ah.mathworks.com>... > Hello, > I work with algorithm to detect cracks or damages on material. I have a lot of images with and without damage (cracks), so do you have any ideas how do I detect a cracks on it? I have thought about this problem a lot, but never was motivated enough to do anything about it. We do have a paper on a fairly elaborate method "Symbolic Dynamic Filtering for Image Analysis: Theory and Experimental Validation," Signal, Image, and Video Processing, Volume 4, No. 3
I'm hardly an image processing expert, but this simple method looks like it might work: 1. Register the image so that the same spot on the sample corresponds to the same pixels in the image. 2. Correct the light levels so they are all the same (difficult because the cracks will lower average brightness levels). Histograms? 3. subtract an image from the uncracked specimen. 4. Cracked areas show large differences. You can use some sort of metric to calculate your state of damage.



