基于深度学习的细纱断头检测模型  

Broken Yarn Detection Model of Spinning Frame Based on Deep Learning

作　　者：吴旭东 吕汉明[1] Wu Xudong;Lv Hanming(School of Textile Science and Engineering,Tiangong University,Tianjin 300387,China)

机构地区：[1]天津工业大学纺织科学与工程学院,天津300387

出　　处：《天津纺织科技》2020年第2期42-47,共6页Tianjin Textile Science & Technology

基　　金：国家自然科学基金(51005168)。

摘　　要：传统的图像处理方法不能准确地对环锭纺细纱机细纱断头进行检测,而经典的深度学习模型直接用于细纱断头的检测其效果也不甚理想,针对这些问题,提出一种基于卷积神经网络的DBYC分类模型,该模型包括改进深度可分离卷积核、提出通道交换操作、添加局部响应值归一化层和使用Dropout层。最后,将模型部署在CPU处理器上。试验结果表明,该模型具有更好的识别效果和更强的泛化能力。Traditional image processing methods can’t accurately detect the broken yarn of ring spinning frame,the classical deep learning model also can’t detect them directly well. To solve these problems, this paper proposes a DBYC classification model based on the convolutional neural networks. This model modifys the depthwise separable convolution, proposes a channel exchange operation, adds a local response normalization layer and uses a dropout layer. Finally, this model is deployed on the CPU processor. The experimental results show that this model has better recognition effect and stronger generalization ability.

关 键 词：细纱断头检测 计算机视觉 深度学习 卷积神经网络 图像分类 分类模型 

分 类 号：TS111.8[轻工技术与工程—纺织材料与纺织品设计]