A novel feedback error-correcting algorithm for automatic recognition of the color and weave pattern of yarn-dyed fabrics

Abstract

In this paper we propose a highly efficient novel algorithm named Automatic Feedback Error-Correcting Color-Weave Pattern Recognition algorithm (AFEC algorithm). This algorithm is capable of simultaneously recognizing the color and weave pattern of yarn dyed fabric. The AFEC algorithm consists of three main components: 1. Color pattern recognition, 2. Weave pattern recognition, and 3. Feedback error correction. The last two each include one additional sub-algorithm, namely, the Infill algorithm and the Rectification algorithm. The first component focuses mainly on using the simplest method to reduce the time demand of color pattern recognition. With this objective we have adopted an X-means clustering algorithm which has less time complexity than other common algorithms used in this field. Furthermore, since the detection of the yarn edge and color classification in this component are designed to be independent of one another, to save time they can be run in parallel. In the second component weave pattern is detected, based on the color pattern obtained in the first component. In the second component the Infill algorithm can identify the logic embedded in the incomplete weave pattern, and hence fill in the gaps to form a complete pattern. By contrasting the color and weave patterns of the fabric, the third and final component uses a Rectification algorithm to correct errors in the recognition of color and weave pattern that may have occurred in the earlier components. Theoretical analysis, and experiments conducted during the present study, indicate that without prior knowledge the AFEC algorithm can improve the accuracy and runtime required for recognizing the color and weave pattern of yarn-dyed fabrics.