Segmentation and Density-Evaluation of Fiber-Reinforced Materials by Dual-Energy Computerized Tomography

Haßler, Ulf; Rizo, P.; Garnero, Line

doi:10.1007/978-1-4613-0383-1_60

Cited by 1 publication

(1 citation statement)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A 14 mm diameter SiC matrix (2.3 g/cm3) containing bundles of C fibers « 20 11m diameter, l.8 g/cm3) perpendicular to the image, and a few holes, was simulated [4]. The fibers VF is 10% in the matrix, 30 to 45% in the bundles.…”

Section: Description Of Simulated Datamentioning

confidence: 99%

Dual-Energy Computed Tomography for Ceramics and Composite Materials

Robert

Dinten

Rizo

1996

Review of Progress in Quantitative Nondestructive Evaluation

Self Cite

View full text Add to dashboard Cite

INTRODUCTIONThis work is part of the Brite-Euram project BRE5535 "DUALETO", whose purpose is to set-up a Dual-Energy Computed Tomography (DE-CT) system suited for the examination of fibers-reinforced composites. DE measurements are obtained by a microfocus X-ray source, and a multi-wire proportional chamber which simultaneously collects projection data at two different energies over a section of the object [1].General data processing is sketched in Figure 1. The pre-reconstruction calibration separates the DE projection data into fibers and matrix components, assuming that their photoelectric (low-energy) and Compton (high-energy) attenuation properties differ significantly [2][3]. Then fibers and matrix images are reconstructed using the filteredbackprojection (FBP) algorithm. These images are quantitative in the sense that the pixel value is really a density (glcm 3 ), but their quality is poor for many reasons. The flux emitted by the source is low, the detector has a low counting rate, and matrix and fibers may have similar compositions. Assuming a 40 Ilm spatial resolution, the structures in the images cannot be individual fibers (7-30 J.Lm diameter), but bundles of fibers, matrix and holes. The low photon flux and the DE calibration process imply considerable noise. Further processing is thus needed. A method based on a Bayesian segmentation, and then a non-linear filtering of the density images according to the segmented image, is proposed. This paper first presents the DE calibration method. Then simulated DE images illustrate the need for post-processing. The third part is dedicated to the Bayesian segmentation. Lastly, we describe how the segmented image can be used to enhance the density images and to compute a map of the volumic fraction (VF) of the fibers.Another research [4] is oriented towards a priori calibration (based on a physical model rather than on the experimental determination of the calibration function) and unsupervised segmentation. PRE-RECONSTRUCTION DUAL-ENERGY CALIBRATIONA CT projection over a polychromatic energy band [EI, E2], for a given ray-sum r, and for an object composed oftwo basis materials "fibers" and "matrix", is:

show abstract

Section: Description Of Simulated Datamentioning

confidence: 99%