Abstract:
|
Thousands of laser confocal microscope views of interior wall surfaces are available as digital images from a single nuclear material container. Do these images have evidence of corrosion events, in particular, potential pitting or potential cracking, that may be of concern for possible thru wall breaching? An analytical methodology is described and implemented in software to rank the images as most likely to least likely to be of concern for these corrosion events. This image ranking may be used by a corrosion subject matter expert to order images for further human visual analysis, possibly aided by certain analytical images that are formed as part of the detection methodology. This methodology augments previous work that focused on the identification of potential pitting by additionally handling the identification of potential cracks.
|