The Research Of Fabric Defect Detection Based On Adaptive Wavelet

Defect inspection is a vital step for quality assurance in fabric production. For a long time, fabric defect detection is primarily performed by human vision, by which there is high rate of missed detection, low efficiency and low reliability. That manual defect inspection ratio only is 40% -60% is indicated based on statistic. With the development of computer and digital image processing technology,computer vision technology has been increasingly applied to detect and classify fabric defects automatically to replace the hand-counting method. But how to improve the efficiency and reliability of fabric inspection has been a focal point in fabric inspection research, and it remains challenging. Wavelet analysis, which has the characteristic of preserving and exhibiting a locality and multi-scale of position-frequency representation for analyzing localized features on an image, often be called as "mathematical microscope". Therefore, scholars pay high attention to the research of fabric defect segmentation based on wavelet, which has been proven to play an important role in defect inspection.By analyzing wavelet transform theory and methods of fabric defect segmentation based on wavelet in depth, approaches of automatically segmenting fabric defect based on adaptive wavelet are proposed, which include: (1) In view of features of different textures distribution between a defect-free fabric and a defective fabric, according to the restrictions of orthogonal wavelet, fabric texture feature is extracted, and the fabric defect segmentation based on an adaptive wavelet is performed by a target function of average energy ratio between them; (2) Aiming at edge segmentation of fabric defect, a design criteria of "best" edge detection filter is researched and analyzed, a approach of fabric defect inspection based on an adaptive bi-orthogonal wavelet is presented by choosing a optimal target function of maximal average modulus ratio between the edge and non-edge. The experimental results show that these approaches have a good ability to detect fabric defect and robustness, compared with the existing methods. Finally, this paper summarizes done works, and gives the recommendations on the future research direction in brief.