基于图像处理的纺织品耐摩擦色牢度评级

为解决目前纺织品色牢度人工评级方式的主观性和繁重工作量，结合深度学习与传统纺织检测，以纺织品检测中的纺织品色牢度评级为对象，研究基于图像处理和深度学习的智能评级创新方法。针对场景与问题，选择利用图像处理技术进行采样图像的预处理和分割，在小样本、多分类的实际条件下搭建数据库，利用深度学习完成对摩擦沾色试样色牢度的迅速评级。结果表明，所选择的图像处理技术对图像的处理效果良好，对后续深度学习准确率的提高有辅助效果；深度学习对耐摩擦沾色试样色牢度的评级准确率达到87.5%,高效、客观且准确率高，实现评级操作简易化，利用神经网络达到代替人眼评级过程，提高准确度和改善目前方法的不足。

Evaluation of textile color rubbing fastness based on image processing

In order to objectively evaluate textile color fastness and eliminate the heavy workload in manual color evaluation, this paper reports on research into evaluation of textile color rubbing fastness using image processing technology. In view of the scenes and problems, image processing technology was used to treat and segment the sampled images. A database was built for small samples from multiple categories, and deep learning was used to achieve the rapid rating of the color fastness of rubbed samples which was necessary for stain removal. The results show that the selected image processing technique works well and has an auxiliary effect on the subsequent improvement of the accuracy of deep learning. The rating accuracy reaches 87.5%, which is efficient, objective and accurate. The rating operation was simplified, and the neural network was used to replace the human eye rating process, improving the accuracy and overcoming the shortcomings of the current method.