Research on digital intelligent recognition method of the weave pattern of fabric based on the redundant information

Digital intelligent recognition for the weave pattern of fabric plays an important role to improve automation and artificial intelligence in textile production process. In order to improve the data processing efficiency and minimize the negative influence such as human error in the conventional methods, a rapid, automatic and accurate method for the surface structure analysis and the fabric weave pattern recognition is proposed. First of all, an imaging system was designed to obtain the double-faced images of fabric samples, and then the captured images were treated by projection algorithm in both warp and weft directions to generate a grid net which splits the image into massive nodes. In the following step, the nodes were preliminary classified based on the intensity of the node’s quadrilateral boundary and at the same time, the color of the nodes was calculated by using the color clustering method. To improve the accuracy of node classification, the types and color information of the adjacent nodes, together with double-faced image information, were utilized for error correction. At last, the node information acquired was encoded and expressed digitally by a basic matrix, two one-dimension matrices (row and column) and a color mapping table. Following the procedure above, the digital model of the weave pattern of the sample fabric is established. Experiments have been conducted and show the performance of the proposed method.