Research On Key Algorithms Used In Automatic Recognition Of Woven Fabric Structures

Nowadays, in the textile industry, the recognition of woven fabric mainly relies on manual work. This method leads to extremely low efficiency although it can ensure high accuracy, so it cannot meet the current high demand of textile industry automation. Applying the computer vision and digital image processing technology to the recognition of woven fabric structures can improve the efficiency of textile industry greatly, which have great significance on automation of textile enterprise.This paper mainly research on key algorithms used in automatic recognition of woven fabric structures based on image processing. We realize the following functions:the automatic location and pretreatment of woven fabric structure blocks, the segmentation of yarns and location of fabric points, the feature extraction of fabric points based on texture and color, the classification of fabric points and the automatic generation of pattern grid.The introduction of maximization thought can locate different woven fabric blocks from whole image with several structure areas. Then, the pretreatment should be executed including Gaussian Blur, histogram equalization and Hough transform.To segment yarns and locate the fabric points, we analyze the accumulated gray curve to determine the clearances and density of yarns. Meanwhile, import the morphological method to solve the problem of highlighting interference.To extract the fabric points’features, we use GLCM and color moments to extract textural features and color features. Through K-means clustering, the feature vectors can be divided into two classes. Furthermore, the correlation parameter can be researched to determine which fabric points are formed by warp yarns and which are formed by weft yarns. Hereby, the system can generate the pattern grid.The experimental results show that this system can identify the basic fabric structures correctly, and it show great theoretical significance and application value.