Year
2004
Abstract
An area of current interest in homeland defense and nuclear safeguards is the application of handheld gamma-ray spectrometers to radioisotope identification. The primary intended use is the secondary screening of containers, persons, or vehicles after an alarm from a portal monitor. A difficulty for existing instruments using peak identification methods is that distortion of the gamma spectrum caused by intervening materials can make isotope identification unreliable. Material basis set (MBS) attenuation-correction algorithms can significantly improve isotope identification with attenuated spectra. However, the MBS method requires repeatedly solving an N×M linear system of equations for each isotope in a library, where N is the number of isotopes and M is the number of spectrum channels used. To speed this process, we have developed a method called the precomputed non-negative least squares (PNNLS) algorithm. This method, which for N less than five is feasible for use in handheld instruments, has been implemented using precomputed pseudo inverses from a singular value decomposition. The performance of the method is discussed and compared for various computational platforms. In addition, we present a fast PNNLS method for solving for the best two isotopes in ordinary response function fitting.