AUTOMATIC CHARACTERIZATION OF NEEDLEPUNCHED FELTS BY CONTENT BASED IMAGE RETRIEVAL - CBIR
BELINO Nuno, NUNES Mário, FIADEIRO Paulo, Alexandre António Da Costa Luís · DOAJ (DOAJ: Directory of Open Access Journals) · 2018
The needled felts whose fibrous bats had been gotten by the classic carding process give us fabrics in which staple fibres follow a marked preferential direction of the machine workflow. This direction is named, as machine direction, main direction or MD and the perpendicular direction is known as cross direction or CD. In consequence of this fibre deposition, the mechanical properties of these products are different when submitted to longitudinal or transversal loads, with a much higher strength in the CD direction and a greater elongation in the MD direction. However, some technical applications of the nonwoven industry, such as filters, geotextiles, and some felts for the automotive industry demands equal mechanical properties along all their directions, particularly, an even MD:CD ratio. In order to introduce additional control in the production system and to achieve a final fabric with the desired behaviour, we envisioned an an automatic system for the pre-needling web drafting control. The management of the pre-needled felts drafting operation was made by an automatic inspection system based upon textural descriptors and content-based image retrieval, which assured optimal drafting conditions to attain a quasi-isotropic fiber distribution and an even MD:CD ratio. Our findings have shown that the comparison between collected images and reference images of the same product stored in the database - for the same feature vector and metric distance – is an excellent tool for the remote and continuous supervising of needlepunched felts quality.