Research On Defect Detection Algorithm For Lace Fabric Based On Image Processing

Lace fabric also called drawnwork or lace,refers to a patterned and banded fabric for decoration.Although the production history of our country later than Eu-rope,but the level of production technology improved very quickly,a large scale of production had been formed.In the productive process of textile,defect detection plays an important role in improving the quality of fabric.At present,in our country textile mainly relies on the manual to carry on the quality inspection,the inspection process is easily affected by human factors,the wrong inspection situation often oc-curs.In order to reduce production costs and improve the accuracy rate of detection,manual testing should be replaced by automatic defect detection.The purpose of this paper is to design the algorithm of automatic defect detec-tion of lace fabric.Lace fabric as the main research object,defect detection algorithm is mainly divided into three steps:image preprocessing,feature extraction of fabric,defect detection.Image preprocessing includes image binarization and image denoising,as the environment of fabric production is complex,there is a problem of uneven bright-ness in the image of the lace,which brings great difficulty to the next step feature extraction.This paper focuses on the research of gray image binarization method,presenting a new binarization method based on Otsu's method,by using the tradi-tional Otsu's method slove the threshold under small rectangular mesh image,and then in the triangle mesh,the threshold is smoothed,finally a more ideal binary im-age is obtained.And then in the triangle mesh,the threshold is smoothed,obtained a more ideal image binarization.Then the median value filter is used to denoise the binary image,and the noise of the patterned texture in the binary image is removed,getting ready for the next step of feature extraction.In the fabric,the patterned texture is cyclical,there is a minimum size of the entire region,the patterned texture can be obtained through the translational of this area,the minimum area is called a lattice.A lattice can be decomposed into a finer component called motif,the lattice can be obtained though the translational,rota-tional,reflectional etc operation of a motif.During the feature extraction of fabric,draw lessons from the method based on local feature in feature extraction of face recognition method,the basic idea is using local geometric properties of pattern tex-ture to find the key points for each motif.Then the key points are used to block the lace fabric,getting the lattice and motif image of the lace fabric.The third step is defect detection,using the method of the energy of moving and variance by Ngan et al.This method can ignore the motif image slight deforma-tion and misalignment,and the energy of moving amplify the defect information of defective motif,make the defection easier to be detected.By determining the range of the energy and variance of arbitrary two motifs of a lattice,determining whether there is a defect in the lattice image.If the calculation results of energy and vari-ance are within the range of determination,the image is defect free,otherwise it is defective.This method can judge pattern confusion,oil and other large defects,in order to further determine the smaller defects such as holes,scratches,etc.,using the above method,defect detection for the sub window image from the motif.Experiments were performed on 217 existing images,the results show that the algorithm is feasible.