Oil Exploration
Exploration is very important for resource extraction, like oil extraction. But many natural resources lie in layers of the earth which is difficult to detect directly. Fortunately, CT scan technology can show the inside of the rocks.
CT image segmentation is an important step of CT image analysis of natural rocks. Gray scale CT volumes need to be segmented into a discrete form such that we can perform quantitative characterization of pore space features.
There are many image segmentation methods for CT images of rocks based on different algorithms. However, most algorithms cannot be efficient and accurate at the same time. For example, the entropy-assisted kriging method is a typical way that can generate high-accuracy segmentation results, but it needs about 12 minutes to process one image in average.
On the other hand, the technology of machine learning for image segmentation processing has been sufficiently mature U-net is the most successful CNN architecture for image segmentation in recent years. So, image segmentation processing with machine learning technology, like U-net, may help us analyze CT images of rocks with higher accuracy and shorter time.