Auxiliary Fabric Defect Detection Algorithm Based On Fuzzy Clustering And Pattern Version

Face with increasingly intense competition of global textile market, improving the quality of textile increasingly and reducing the cost of textile become an effective means of holding an invincible position for the textile enterprises. During a series of technological process of production of fabric, fabric defect does not only decrease the quality of fabric, but also cause huge waste of the material, human and material resources. Thus, how to reduce and avoid fabric defect becomes an important subject of improving the competitiveness of textile. Computer aided detection in fabric defect becomes an important and effective means.So far as the status, most efforts of fabric defect detection aim at the defect that is caused by loom or other relative reasons, which is called weaving defect in this thesis. The defect(which is called weave defect in this thesis) which is caused by jacquard fault happens frequently, such as electromagnetic interference between magnets, lift-ing needle deformation or magnetization, and so on. The essence of weave defect is the property of the interlacing point of weave have deviated from the interlacing point that is designed.In this thesis, weaving defect and weave defect are studied deeply, and two dif-ferent defect detection algorithms are proposed, which aim at different characteristics of weaving defect and weave defect. An detection algorithm based on fuzzy clustering is proposed for the weaving defect. The gray average projections in the weft and warp direction are considered to be eigenvalue in this algorithm. After the eigenvalue is extracted, the suspicious defect regions, which are called pseudo defect regions in this thesis, are separated using F-statistics fuzzy clustering algorithm according to the pro-jection distortion phenomenon of gray average in defect region. The defect regions are located and extracted from suspicious defect regions by selecting proper density and distortion thresholds so that the pseudo defect regions are filtered out effectively. The algorithm is verified and analyzed in the experiment part by means of common weave defects at last. The results of experiments show that the algorithm is not only easy and feasible, but also suitable for most of weave defect. A defect detection algorithm aided by jacquard cards is proposed for weave defect. The fabric image is divided into each weft unit according to the fluctuation pattern of gray average projection in the weft direction. The properties of intersection point of each weft unit are identified re-spectively. The weave defects are identified by comparing the property of intersection points with corresponding jacquard cards. The results of experiments show that the algorithm have the features of high reliability, strong adaptability and real-time.