An optimised method for material identification using a photon counting detector

X-ray photon counting detectors have become a competitive alternative to energy integrating systems in certain domains. However, processing methods currently used in radiography for investigating the composition of an object are an extension of dual energy methods and have started to show their limits with an increasing number of counting windows (bins). Thus, in a context of material recognition for homeland security, this study aims to introduce a new data processing method suitable for any type of detector, in integrating or counting mode, regardless of the number of bins. Additionally, a criterion to quantify the influence of the number of counting windows as well as the influence of their threshold position has been developed.The optimisation process is conducted in simulation by considering ideal detectors, and the results, for detectors with m=2, 3, 4 and 5 counting windows, are compared to those obtained with an analytical criterion developed in the literature and adapted to our study context. Both methods give identical results. In the final part, the performances of a spectrometric detector (energy windows width of 1 keV) and of optimised systems with m=2, 3, 4 and 5 bins are analysed for homogeneous plastic identification. The results show an increase in performance with increase in bin number until the performance level of the spectrometric detector is reached.Moreover, a discussion on the optimisation robustness as a function of material thickness to be identified is presented. The spectrometric detector, which does not require any optimisation of the bins thresholds, appears then to be a candidate of choice for material recognition when using X-ray photon counting detectors.