Many tasks in security, medical imaging, materials science, and non-destructive testing require spatially-resolved material identification and/or quantification. X-ray diffraction imaging can accomplish this task, but it has typically been slow or of too course resolution to adequately address the real world need. In this talk, I discuss our computational approach to X-ray diffraction imaging that allows for 2D and 3D X-ray diffraction imaging at speeds relevant to the task at hand. Through a combination of physical coding, model-based reconstruction, inclusion of side-information, and machine learning-based processing, we demonstrate the ability to evaluate the contents of baggage and parcels in seconds as well as perform high resolution material identification of biological samples on the order of a few minutes using conventional, off-the-shelf components.
|