基于无抽样离散小波变换的复杂纹理织物疵点检测  

Research on defect detection of complex textured fabric based on non-sampling discrete wavelet transform

作　　者：杨晓波 YANG Xiaobo(Zhejiang Shuren University,Hangzhou,Zhejiang 310015,China)

机构地区：[1]浙江树人学院,浙江杭州310015

出　　处：《毛纺科技》2024年第2期133-138,共6页Wool Textile Journal

基　　金：浙江省自然科学基金项目(Y1110023)。

摘　　要：为了进一步提高复杂纹理织物的疵点识别率,本文采用新型算法检测复杂纹理织物的疵点。首先分析了无抽样离散小波的变换原理,选取二维无抽样小波对织物疵点进行检测;接着分析了小波基和分解尺度的选择依据,并在此基础上提出织物疵点的判别流程;最后为了验证无抽样离散小波变换算法的有效性,与其他主流算法进行对比分析。由于抽样离散小波具有平移不变特性,在疵点区域小波变换对应的能量会增大,而在无疵点区域能量会减小,选用Daubechies D2小波作为小波基,小波分解尺度的选择需要考虑织物图像的纹理特征,选择的尺度以适中为宜;从织物图像区域中提取水平、垂直和对角线方向能量作为特征值,分别选取6种类型的织物疵点进行对比实验。实验结果表明,采用无抽样离散小波变换算法进行织物疵点检测平均正确率和实时检测速度均高于其他主流算法,可以较好地用于复杂纹理的织物疵点检测。In order to improve the defect recognition rate of complex textured fabric,a new algorithm was used to detect the defects of complex textured fabric.Firstly,the transform principle of unsampled discrete wavelet was analyzed,and two-dimensional unsampled wavelet was selected to detect fabric defects.Then,the selection basis of wavelet basis and decomposition scale was analyzed,and on this basis,the identification process of fabric defects was proposed.Finally,in order to verify the effectiveness of the non-sampling discrete wavelet transform algorithm,other mainstream algorithms were used for comparative analysis.Because the sampled discrete wavelet has translation invariant characteristics,the energy corresponding to the wavelet transform will increase in the defect region,while the energy will decrease in the non-defect region.Daubechies D2 wavelet was selected as the wavelet basis.The selection of wavelet decomposition scale should consider the texture characteristics of the fabric image,and the selected scale should be moderate.Three kinds of energy in horizontal,vertical and diagonal directions,were extracted from the fabric image region as characteristic values,and six types of fabric defects were selected respectively for comparative experiments.The experimental results show that the average accuracy and real-time detection speed of fabric defects detection using the non-sampling discrete wavelet transform algorithm are higher than other mainstream algorithms,and it can be better used for fabric defects detection with complex textures.

关 键 词：无抽样离散小波 织物疵点 特征提取 复杂纹理织物 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]