This is a rather complex problem, and there is no plug-in for this. Using light (structured or laser) or shadow to determine a height of 0.2 mm will almost certainly not work with an acceptable degree of certainty, no matter how many βphotosβ you take. (This is just my personal intuition, in computer vision we check whether something really works).
GLCM is a good feature for describing textures, but as far as I know, it is used to check for the presence of a pattern in the texture, so I believe that this will lead to a positive value for the text in 2D printing, if it is a kind of repeating pattern.
I would let the computer know what text is, what texture is. Just extract large amounts of 3D and 2D data and use the machine learning engine to find out what it is. If the space of possibilities is rich enough, it can find a way to distinguish one from the other, no matter how our human mind can. The function space must consist of edge and color functions.
If the system environment is stable and controlled, this approach will work especially well, since the training data will be so similar to the test data.
For this problem, I will start by calculating the color and edge functions (the sum of the pixels of the local image over different red and color channels) and try the advanced classifier. Accelerated classifiers are not up-to-date when it comes to machine learning, but they don't overdo it well (which means you can just paste in as much data as you want) and are more likely to work in a stable environment.
Hope this helps,
Good luck.
source share