This study focuses on the threshold segmentation algorithm to obtain the real microstructure of asphalt concrete based on digital image technique, the perlite powder which was a kind of low-density material was put in the asphalt concrete to enhance the density contrast, three different specimens in which added different contents of perlite powder were compacted, and then the asphalt concrete specimens were scanned using x-ray CT to capture the gray images that reflect the density differences of the three constituents such as aggregates, mastic and voids, the CT images were converted to be the histograms. Furthermore, the FCM (Fuzzy C-Means) was demonstrated that it could be utilized to choose proper threshold values and segment images exactly, according to the double peak conditions of the three different histograms, the double peak condition for AC-13 is the best among the three types, a similar double peak features between AK13 and SMA-13 were observed. The results shows that the different contents of perlite powder added in the asphalt concrete can form different double peaks. This is another new method to segment the three constituents of the asphalt concrete exactly.
COVSeg‐NET
: A deep convolution neural network for
COVID
‐19 lung
CT
image segmentation
