关键词: 机器视觉, 拉伸性能, 图像处理, 动态形变, 特征提取

Abstract: The application of textile materials has gradually expanded into fields such as smart wearables, construction, aerospace and healthcare, posing higher demands on the analysis of basic mechanical properties of materials. Existing measurement methods primarily focus on the overall tensile results of fabrics. However, in the context of evaluating new material developments, there are still many obstacles in distinguishing the localized stress variations in textile materials with different characteristics and treatment methods. How to fully utilize the advantages of non-contact measurement to accurately capture subtle changes during the fabric tensile process and how to establish a reasonable connection between these changes and the mechanical properties of the fabric remain issues that require in-depth research.
This paper proposed a method to explore the relationship between fabrics' dynamic tensile stress and morphological changes based on machine vision technology. Through comparative research with fabric tensile tests, the feasibility of applying machine vision technology in fabric tensile tests was explored, and the consistency of results between machine vision analysis and traditional tests was demonstrated. A self-built video sampling device was used to record the stretching process, with the video frames decomposed into sequential images for steps such as image preprocessing, object segmentation, and feature extraction. The external morphological features of the fabric affected by tensile load were extracted from the images. Furthermore, the Poisson's ratio of the fabric was calculated, and the results were correlated with its tensile properties to expand the methods for detecting fabric tensile performance. Additionally, the feasibility of machine vision technology in measuring fabric tensile deformation characteristic index under different treatment methods was validated.
The results indicate that the data obtained using machine vision technology are in good agreement with those acquired through traditional measurement methods, confirming the effectiveness of the proposed method in this study. Furthermore, to explore the capability of machine vision technology in capturing subtle changes during fabric stretching and its potential for application in fabric performance evaluation, ANOVA or non-parametric tests were conducted on fabric tensile deformation characteristic index under different treatment methods. The research results demonstrate that there are significant differences in elongation among woven polyester, knitted polyester, and woven wool, indicating a clear distinction in the distribution of their elongation characteristics. Moreover, different treatment methods also have an impact on the elongation properties of the materials. Materials subjected to soaking in clear water, one-time washing, and five-time washing exhibit relatively concentrated elongation data with low variability, whereas samples soaked in detergent or subjected to light aging treatment show greater variability. For the shrinkage index at the narrowest point, woven wool and knitted wool also exhibit significant differences. Further analysis reveals that under various treatment conditions, samples subjected to five washes demonstrate higher variability, while those soaked in laundry detergent or exposed to light aging for 30 hours exhibit lower variability. This verifies that machine vision technology demonstrates good separability in fabric characteristic analysis, particularly in the elongation index, where its performance surpasses that of the narrowest point shrinkage index.

Key words: machine vision, tensile property, image processing, dynamic deformation, feature extraction