A novel image processing technology for recognizing the weave of fabrics

The current analysis of fabric weave diagrams requires using fabric analyzing glass to record the weave number manually. This method damages eyesight and is also very time consuming. In addition, the unweaving mode damages the weave structure of woven fabric. This study uses a computer vision system and digital image processing technology for direct non-destructive analysis of the commonly used 12 fabric textures of woven fabrics without unweaving. Moreover, it proposes an automated woven fabric weave recognition method to enhance the practicability and fault tolerance of the recognition system. Firstly, the woven fabric image was shot by using a front light source and back light source, the noise of the woven fabric image was reduced by using a median filter and the contrast was increased by using histogram equalization. The statistical threshold value was used to segment the warp yarn area and the opening operation of morphology was used to disconnect the connected blocks and erode small noise. Horizontal projection and vertical projection were used to segment the warp yarn and weft yarn. The weave diagram was drawn to improve the computing time of the gray-level co-occurrence matrix. The contrast in the gray-level co-occurrence matrix was selected as the eigenvalue. In terms of woven fabric samples, 12 target samples were obtained, the Euclidean distance classifier was used and the 12 test samples were used for the experiment. The result showed a recognition rate of 100%. The recognition system was adopted by this study to effectively recognize the woven fabric weave.