A Study Regarding To Auto-correction And Organizational Analysis For Woven Fabric Images

With widely application of computer technology, production automation has become an inevitable trend during the development of the textile industry. However, nowadays, fabric texture's analysis and identification mainly rely on human experience or professional tools, and this traditional approach has become the bottleneck of textiles'rapid design and production. Therefore, in order to replace artificially detection, it is both theoretically and practically significant to develop an effective system which can detect and analyze fabric material automatically. Based on this background, this paper presented a study regarding to auto-correction and organizational analysis for woven fabric images. The research points and achievements are listed as follows:1. The research outlined an approach which used image processing techniques to conduct an automatic parametric analysis and organization structure identification for fabric material, and also addressed an overall proposal concerning fabric texture analysis and automatic identification.2. In the field of fabric image pre-processing, this paper proposed a pre-processing algorithm suitable for fabric images, as well as its implementation measures. And meanwhile, a fabric image rapid skew detection and correction algorithms based on the improved Hough transform was presented. By combining the weft direction information extracted by row-difference operation, this method performed hierarchical Hough transform on the weft boundary to detect the skew angle of fabric image. Eventually, a rotation algorithm based on the image linear storage structure was introduced, and the skew image was corrected rapidly.3. In the part of fabric image parameter extraction and analysis, a wavelet transform based algorithm to measure the warp and weft density, and the yarn fineness of fabric was presented. After pre-processing, the two-dimensional discrete wavelet decomposition was applied on the fabric image, and binaryzation and smoothing techniques were adopted on the sub-image of warp and weft directions respectively, then the yarn texture can be separated from the fabric image. Finally, we can obtain the warp and weft density and the yarn fineness via calculations.4. In the study of the fabric organizational structure segmentation and identification, a fabric image segmentation method based on multi-scale Markov random field (MRF) was presented. It combined with different scales of the edge information, which is extracted by the modulus maximum of wavelet transform, and used the Markov random field segmentation algorithm to segment the fabric image, then through the organization point edge leveling and identification to generate a fabric weave diagram.This paper incorporated professional textile knowledge and computer image processing techniques, to provide the corresponding algorithm for key issues in the organization analysis system based on a woven fabrics image. It not only effectively solved those problems including the fabric image skew detection and correction, yarn density and fineness test, organization structure segmentation and identification and so forth, but also improved the automotive and intelligent level in the fabric organization analysis system. This study has an important practical significance for the promotion of textile production automation and the development of the textile industry.