Neutron scattering correction functions for neutron radiographic images.

Abstract

Neutron radiography is a valuable nondestructive testing tool to image objects invisible to X-rays, because neutrons interact with the nucleus of an atom while X-ray interacts with the electrons. While the absorption and scattering of neutrons through the object materials are the key phenomena which provide the desired information on the materials, neutrons scattered off the beam direction degrade the image resolution. We have developed an image reconstruction algorithm that involves the point spread functions (PSFs), to account for image degradations due to neutron scattering and due to image system unsharpness. The algorithm is based on modulation transfer function approach and uses a maximum likelihood method that iteratively estimates the scattering effects and hence the unknown geometry. Inherent fluctuations in the experimental data are represented through a Kalman filtering algorithm. We represent neutron scattering through the PSF and line spread function (LSF), which are analytically obtained from the 2-D neutron transport equation for a monoenergetic neutron beam. Through a successive collision approach, the scattering PSF and LSF account fully for multiple scattering of neutrons. MCNP Monte Carlo calculations have been performed to benchmark the analytic transport solutions. A number of neutron radiographs have been obtained with collimated thermal neutron at the Vertical Beam Pon of the Ford Nuclear Reactor and the images have been reconstructed through our scattering and system PSFs explicitly utilized. While the system PSF is obtained through a knife-edge experiment, we obtained the scattering PSF from the analytical LSFs for an infinite-medium, half-space, and finite slab. The scattering correction algorithm determines neutron scattering components based on the best estimates of the scattering functions and removes the scattering effects in reconstructing neutron radiographic images. We note clear enhancement in the reconstructed images in an idealized experiment performed with two paraffin disks with different radii overlaid. In this test, we are able to significantly improve our estimate of the paraffin thicknesses obtained from the actual neutron radiographs. Combined with an iterative reconstruction technique, the scatter-correction algorithm provides a practical method to perform quantitative evaluation of radiographic images without sacrificing the neutron intensity on the image plane.