A scatter correction method for Multi-MeV Flash Radiography

TL;DR

This paper introduces a new online scatter correction method for Multi-MeV flash radiography, improving density measurement accuracy in high-density objects through collimation and advanced Monte Carlo simulations.

Contribution

The paper presents a novel scatter estimation technique using collimation and an accelerated Monte Carlo method for improved density reconstruction in high-energy radiography.

Findings

Evaluation error of scatter less than 2% at high densities

Reconstructed density error below 2% in simulations

Method does not require prior knowledge or regularization

Abstract

Multi-MeV flash radiography is often used as the primary diagnostic technique for high energy and density (HED) physics experiments. Primary X-ray which is attenuated by the object offers density information of the object. For a thick metal object with area density as high as 150 g/cm2, the rest part of primary X-ray which passes through the object may drowned in scattered X-ray fog. It seriously limits accuracy of density quantification. In this research, an online scatter estimation method is newly designed which can be easily arranged by putting an additional slit collimator downstream of the general X-ray radiography layout. The basic ideal of this method is that the proportion of scatter and primary x-ray will be changed a lot when x-ray passes through a slit like collimation, then scatter component is solvable with known the collimation performance of the silt on scatter and primary x-ray. Monte Carlo simulation shows that, with this method, evaluation error of average scatter is less than 2% when object area density is as high as 200 g/cm2. In addition to the average scatter, an accelerated Monte Carlo method is developed to obtain scatter distribution in iterative reconstruction. By employing the Genetic algorithm as an optimizer, reconstruction can be done by searching a density which projection with scatter best matches experimental result. This reconstruction method requires neither a priori-knowledge such as mass restriction nor regularization. Simulations show that for France Test Object (FTO), the error of reconstructed density is less than 2%, and uncertainty basically covers the real density.