Neutronography, or neutron radiography, consists of producing a 2D image of an object crossed by a neutron flux by measuring the differences in absorption and scattering of these particles as they pass through the materials. It is a non-destructive diagnostic and these images have extremely interesting characteristics, very different from those obtained by X-ray. Indeed, neutrons, which are essentially sensitive to the nuclear interaction, are affected by light chemical elements (hydrogen), present in particular in organic materials, whereas most heavier elements, such as metals, are transparent to them. Neutronography thus finds unique applications in materials science, engineering, archaeology or the study of works of art.
Until now, nuclear research reactors have been producing neutrons, but these facilities are at the end of their lives, such as the Orphée reactor in Saclay, which was shut down in 2019. Alternative sources are developed, based on neutrons emitted during nuclear reactions produced by a beam of accelerated particles (e.g. protons), such as the SONATE project. These new facilities are cheaper and more flexible than nuclear reactors, but provide lower neutron fluxes. To avoid excessively long measurement times, it is necessary to use imaging technologies that are more sensitive than traditional silver films. The aim of this thesis is to qualify different modern imaging technologies and to optimise them for industrial neutronography. Different detector technologies are possible: detectors based on microchannel wafers, or detectors based on scintillating films coupled with CCD cameras. The analysis of the signals from these detectors can also be optimised, and the "event by event" mode can be studied to improve the selectivity and resolution of the images. Finally, post-processing of the image, based on noise reduction and super-resolution algorithms, can also be used, which can make use of advanced machine learning methods for image reconstruction, potentially allowing a gain in reconstruction quality as well as rapid analysis.