Developing Intelligent Radiation Analysis Systems: A Hybrid Wave-Fuzzy Methodology for Analysis of Radiation Spectra

Analysis of measured radiation spectra is performed using specialized algorithms developed to identify isotopic signature patterns that indicate the presence of specific radioisotopes in the presence of a radiation background signal. In this paper an intelligent methodology for detection of signature patterns in measured radiation spectra is presented. In particular, the methodology performs signature detection by implementing two computational tools: i) wavelet processing, and ii) fuzzy logic inference. Initially, wavelet processing is applied to identify spectral maxima in the measured signal, subsequently a fuzzy logic inference engine is employed to match detected maxima to known entries in a isotope peak library. The library is comprised of entries that represent the known energy of characteristic peaks for specific isotopic signatures. The presented methodology is benchmarked against a multiple linear regression (MLR) spectrum fitting algorithm applied to a set of synthesized gamma-ray spectra taken with a low resolution NaI radiation detector while the presented methodology outperforms MLR in the vast majority of the tests by providing correct detections with high confidence, and a low number of false detections.