Working While We Wait

Posted by Ethan Hilton on October 15, 2015

Due to unforeseen delays in collecting actual data, we decided we needed to press on as best we could without having actual images to begin evaluating using 2-point statistics on. It therefore became necessary to create “fake” images on which we could begin to perform statistical tests. The hope was that this would streamline the process of executing these statistical tests on images of actual membranes once they were obtained. This time-saving tactic may prove to be crucial as we will have a limited amount of time with the data we were anticipating to get six weeks ago. This post primarily discusses how this “fake” data, or representations, were created. The representations were created using the MATlab script “sample_creator.m” and represented a 2-phase material: a membrane 400 microns thick and 1600 microns long with suspended particles that vary in size from 5 to 22 micron. A resolution of 5 microns was set to simulate the resolution of the CT scanner used to collect the data of the actual membranes. Dividing the length and width of the representation by the resolution factor produces the number of pixels represented in each dimension (M, N). These pixels are formed into a matrix of zeros M by N. The particles are then created to be a random size using the function “particleDiameter.m”. The size of the particle is generated using a size probability provided by the manufacturer of the particles. The particle location is determined using the function “particleLocation” to assign the x and y coordinates of the center of the particle with the origin at the bottom left corner of the representation. The function randomly distributes particles in the x-direction (along the length) and distributes the particle using a probability model based on local shear in the y-direction (along the thickness). The distribution can also be biased towards the top or bottom surface by adjusting the probability variables . The “particleShade” function uses the info generated about the size and location of the particle to determine which components of the matrix representation should be changed to a “1” to represent a separate phase of the material. An example of the resulting sample representation is shown in Fig. 1 below.

Fig 1 Fig.1 - Sample representation of membrane with suspended particles.

All MATlab codes used in this process can be found on GitHub.