Tuesday, November 23, 2010

Spectrum Unfolding and L1 Minimization ?

Nuclear detectors are a very challenging design problem because unlike light detectors featuredin computational photography, all the elements of the detectors do interact with the particles being tracked. This is the case in particular of neutron detectors. As I was reading on the subject ( in Radiation Detection and Measurement ) for the purpose of evaluating a compressive equivalent to current technology, I came across the problem that is well studied in this area called spectrum unfolding, i.e. the fitting of the energy spectrum recorded by detectors as a linear combination of simple spectra of known interaction of the detector with specific particles at specific energy.. This is a situation that is encountered in hyperspectral unmixing: or how can a signal be decomposed as a series of different spectra found in some dictionary. To give more context here is a good introduction on the subject in this presentation entitled: Unfolding techniques for neutron spectrometry by Marcel Reginatto



Of interest is that none of the "newer" methods of investigation include reconstruction solvers used in compressive sneisng:


Surely, L1 minimization would provide good results here as well. The next step would be to design a sensor that permits either a faster way of performing this spectrum unfolding or provide additional spatial information , a feature that is currently not available.More on that later....

Other papers and webpages of interest:

2 comments:

Anonymous said...

"...Nuclear detectors are a very challenging design problem because unlike light detectors featuredin computational photography, all the elements of the detectors do interact with the particles being tracked..."

Hmmm, isn't that the case that also light detectors interact with the particles (photons) being tracked?

Igor said...

The key word in that sentence is "all". The cables, the casing, everything in the detector "reacts" with the radiation. The sensitive part of the detector gets the sum of all these interactions between the whole detector and the radiation being tracked.

Hope this helps,

Cheers,

Igor.

Printfriendly