基于方向灰度积分曲线特征的织物疵点检测算法研究  

Research on the fabric defect detection algorithm based on the featureof directional gray integral curves

作　　者：李丑旦 祝双武 马阿辉 王世豪 马晓彤 LI Choudan;ZHU Shuangwu;MA Ahui;WANG Shihao;MA Xiaotong(School of Textile Science and Engineering,Xi’an Polytechnic University,Xi’an 710048,China)

机构地区：[1]西安工程大学纺织科学与工程学院,西安710048

出　　处：《丝绸》2023年第4期51-60,共10页Journal of Silk

基　　金：陕西省教育厅科研计划项目(18JS042);中国纺织工业联合会科技指导性项目(2019057)。

摘　　要：针对于断纱、缺纱、穿错、粗纱等这类结构型织物疵点,由于其具有灰度跳变不明显、疵点面积小的特征在疵点检测过程中难以检测这一问题,本文结合织物图像自身的纹理特征及结构型疵点的方向性特征,创新性地提出基于方向灰度积分曲线特征的织物疵点检测方法,将二维织物图像的疵点检测转化为对一维灰度积分曲线特征的分类识别。该方法通过对输入的图像提取垂直水平方向灰度积分波形曲线,并对积分曲线提取了包括平均值、方差、能量等14个波形特征,然后利用可优化SVM分类算法对提取特征进行疵点判别。通过对漏针、断纱、并经、粗纱等疵点进行检测试验,结果表明,本文提出的疵点检测方法不仅对检测灰度跳变较小的结构型疵点具有较好的检测效果,检测准确率达到了94.34%,而且检测速度快,可以满足实时检测的速度要求。Textile fabric appearance quality inspection is an important process in textile production and directly affects the quality of textile and product pricing.In the production process of textile enterprises,structural defects such as broken yarn,lack of yarn,wrong threading,roving yarn and other defects are difficult to be accurately detected through the traditional gray-based method due to the smaller gray difference and defect area compared with the normal non-defect area,which is an important difficulty in the development of automatic detection technology for fabric defects.In order to solve the problem that structural fabric defects such as broken yarn,missing yarn,wrong threading and roving are difficult to detect,we analyzed and studied these defects and found that the structural defects showed obvious directional characteristics on the image.By conducting directional gray integral projection on the fabric,we found that the period of fabric gray integral curves would be damaged partially in the defect area,and the defect position on the projection curve had a clear mutation.Based on the above studies,we proposed the fabric defect detection algorithm based on the gray integral curve features.By extracting the characteristic values of the gray integral curve in the fabric direction,the two-dimensional fabric image defect detection was transformed into the classification and recognition of the one-dimensional gray integral curve features.In this paper,the vertical and horizontal direction gray integral waveform curves were extracted from the input image,and 14 waveform features including the average value,variance,and energy were extracted from the integral curves,and then the optimizable SVM classification algorithm was used to classify the extraction features.Through the detection experiments on such defects as drop stitch,broken yarn,parallel warp and roving,we draw the conclusion that the proposed method not only has a good detection effect on the structural defects with small gray scale jump,but also has a f

关 键 词：织物纹理特征 结构型疵点 方向灰度积分投影 曲线特征 可优化SVM 疵点检测 

分 类 号：TS101.914[轻工技术与工程—纺织工程] TP391[轻工技术与工程—纺织科学与工程]