基于浅层卷积特征和双低秩表示的织物疵点检测算法研究

Fabric defect detection algorithm based on shallow CNN feature and double low-rank representation

鉴于低秩表示技术可用于实现织物背景和疵点的有效分类,已被应用于织物疵点检测,但相应检测结果依赖于织物纹理特征的有效表征和低秩表示模型的性能,提出了基于浅层卷积特征和双低秩表示的织物疵点检测算法。采用经过训练的浅层网络提取织物图像的深度特征,构建了特征矩阵;建立双低秩表示模型,实现了疵点与背景的有效分离;采用阈值分割算法对疵点区域进行了定位。实验结果表明,所提出的检测算法能有效地定位疵点区域,其检测精度优于其他方法。

Fabric defect detection plays an important role in the quality control of textile.Traditional detection methods have low detection accuracy and poor adaptability.Low-rank representation model can divide the fabric image into defect-free background and sparse defects,and have proven applicable in fabric defect detection.However,the detection results of these methods depend on the effective characterization of fabric texture features and the performance of low-rank representation model.In this paper,a fabric defect detection method based on shallow CNN feature and double low-rank representation is proposed.The deep feature of fabric image is extracted by the pre-trained shallow network,and the fabric texture feature matrix is constructed.Then the double low-rank representation model is established to separate the defect from the background effectively.Finally,the defect region is located by threshold segmentation algorithm.Experimental results demonstrate that the proposed method can detect the defects with high accuracy,and is better than other methods.