Compressed sensing and defect-based dictionaries for characteristics extraction in mm-Wave non-destructive testing

In ultra-wideband non-destructive testing of large multilayered polymers, data collection and reduction can be achieved by applying compressed sensing techniques. In this work, using effective modelling of possible defects, such as air gaps between layers, we construct defect dictionaries and use them as support data for a signal similarity-based classifier, which will automatically extract the main characteristics of the inspected defect.