Individual fibre separation in 3D fibrous materials imaged by X-ray tomography.

Modelling the physical behaviour of fibrous materials still remains a great challenge because it requires to evaluate the inner structure of the different phases at the phase scale (fibre or matrix) and the at constituent scale (fibre). X-ray computed tomography (CT) imaging can help to characterize and to model these structures, since it allows separating the phases, based on the grey level of CT scans. However, once the fibrous phase has been isolated, automatically separating the fibres from each other is still very challenging. This work aims at proposing a method which allows separating the fibres and localizing the fibre-fibre contacts for various fibres geometries, that is: straight or woven fibres, with circular or non-circular cross sections, in a way that is independent of the fibres orientations. This method uses the local orientation of the structure formed by the fibrous phase and then introduces the misorientation angle. The threshold of this angle is the only parameter required to separate the fibres. This paper investigates the efficiency of the proposed algorithm in various conditions, for instance by changing the image resolution or the fibre tortuosity on synthetic images. Finally, the proposed algorithm is applied to real images or samples made up of synthetic solid fibres.