Abstract:  Focusing on the problems of low efficiency and poor stability of the traditional glass fiber fabric defect classification method, a glass fiber fabric defect classification algorithm based on multi-feature fusion is proposed. Firstly, the median filter is used to preprocess the glass fiber fabric image to remove the detail noise and reduce the influence of background texture. Secondly, Canny edge detection is performed on the pre-processed image, and Hu invariant moment is used to extract the geometric features of defects. Then, the texture features of the image are extracted by scale invariant feature transform (SIFT). After K-means clustering, the bag of words model (BoW) of the glass fiber fabric image is constructed. Finally, the geometric features and texture features are merged and passed into the SVM for training, and the corresponding glass fiber fabric defect classification model is obtained. The experimental results show that the average classification accuracy can reach 97.22 %, which can meet the actual needs of enterprises.

Key words: Hu moment invariant, SIFT feature, bag of words model (BoW), support vector machine(SVM)

摘要： 针对传统的电子布缺陷分类方法效率低，稳定性差的问题，提出了基于多特征融合的电子布缺陷分类算法。首先，使用中值滤波对电子布图像进行预处理，滤除细节噪声，减少背景纹理的影响；其次，对预处理后的图像进行Canny边缘检测，利用Hu不变矩提取缺陷的几何特征；再利用尺度不变特征变换（SIFT）提取图像的纹理特征，使用K-means聚类后，构建电子布图像的词袋模型（BoW）；最后，将几何特征和纹理特征融合，并传入SVM中进行训练，得到相应的电子布缺陷分类模型。实验结果表明，应用多特征融合的方法对电子布缺陷进行分类，其平均准确率可达97.22 %，能够满足企业的实际需求。

关键词: Hu不变矩, SIFT特征, 词袋模型（BoW）, 支持向量机（SVM）