The lightweight worsted fabric made of Sirofil yarn was developed and its textile performance was studied. By means of conventional testing, KES and FAST, it is concluded that with the coordination of proper fabric de...The lightweight worsted fabric made of Sirofil yarn was developed and its textile performance was studied. By means of conventional testing, KES and FAST, it is concluded that with the coordination of proper fabric design and processing technique, the performance of this innovative fabric is superior to that of the conventional lightweight fabric Furthermore, it is predicted from the relevant parameters that its processability in the following suit manufacture is very desirable.展开更多
为满足纺织业内机织印花布瑕疵检测的实时性需求,基于利用回归思想进行检测的单阶段算法模型YOLOv_(3)(you only look once version 3),提出一种改进的机织印花布疵点实时检测方法。通过优化骨干网络结构,引入可变形卷积,提高印花背景...为满足纺织业内机织印花布瑕疵检测的实时性需求,基于利用回归思想进行检测的单阶段算法模型YOLOv_(3)(you only look once version 3),提出一种改进的机织印花布疵点实时检测方法。通过优化骨干网络结构,引入可变形卷积,提高印花背景下模型的瑕疵特征提取能力;设计新的损失函数,提高瑕疵分类和定位的精准度;引入几何中位数剪枝算法,去除深层网络冗余参数,进一步提高系统检测速度。试验结果表明,改进算法的模型在测试集上准确率可达92.02%,检测精度显著提高,每张图片检测平均耗时22.61 ms,满足工厂的实时性要求。展开更多
布匹瑕疵检测是纺织业质量管理的重要环节.在嵌入式设备上实现准确、快速的布匹瑕疵检测能有效降低成本,因而价值巨大.考虑到实际生产中花色布匹瑕疵具有背景复杂、数量差异大、极端长宽比和小瑕疵占比高等结构特性,提出一种基于轻量级...布匹瑕疵检测是纺织业质量管理的重要环节.在嵌入式设备上实现准确、快速的布匹瑕疵检测能有效降低成本,因而价值巨大.考虑到实际生产中花色布匹瑕疵具有背景复杂、数量差异大、极端长宽比和小瑕疵占比高等结构特性,提出一种基于轻量级模型的花色布匹瑕疵检测方法并将其部署在嵌入式设备Raspberry Pi 4B上.首先在一阶段目标检测网络YOLO的基础上用轻量级特征提取网络ShuffleNetV2提取花色布匹瑕疵的特征,以减少网络结构复杂度及参数量,提升检测速度;其次是检测头的解耦合,将分类与定位任务分离,以提升模型收敛速度;此外引入CIoU作为瑕疵位置回归损失函数,提高瑕疵定位准确性.实验结果表明,本文算法在Raspberry Pi 4B上可达8.6 FPS的检测速度,可满足纺织工业应用需求.展开更多
基金This research is one of the national "ninth five" key projects, supported by China Textile University and Shanghai Wool & Flax Research Institute.
文摘The lightweight worsted fabric made of Sirofil yarn was developed and its textile performance was studied. By means of conventional testing, KES and FAST, it is concluded that with the coordination of proper fabric design and processing technique, the performance of this innovative fabric is superior to that of the conventional lightweight fabric Furthermore, it is predicted from the relevant parameters that its processability in the following suit manufacture is very desirable.
文摘为满足纺织业内机织印花布瑕疵检测的实时性需求,基于利用回归思想进行检测的单阶段算法模型YOLOv_(3)(you only look once version 3),提出一种改进的机织印花布疵点实时检测方法。通过优化骨干网络结构,引入可变形卷积,提高印花背景下模型的瑕疵特征提取能力;设计新的损失函数,提高瑕疵分类和定位的精准度;引入几何中位数剪枝算法,去除深层网络冗余参数,进一步提高系统检测速度。试验结果表明,改进算法的模型在测试集上准确率可达92.02%,检测精度显著提高,每张图片检测平均耗时22.61 ms,满足工厂的实时性要求。
文摘布匹瑕疵检测是纺织业质量管理的重要环节.在嵌入式设备上实现准确、快速的布匹瑕疵检测能有效降低成本,因而价值巨大.考虑到实际生产中花色布匹瑕疵具有背景复杂、数量差异大、极端长宽比和小瑕疵占比高等结构特性,提出一种基于轻量级模型的花色布匹瑕疵检测方法并将其部署在嵌入式设备Raspberry Pi 4B上.首先在一阶段目标检测网络YOLO的基础上用轻量级特征提取网络ShuffleNetV2提取花色布匹瑕疵的特征,以减少网络结构复杂度及参数量,提升检测速度;其次是检测头的解耦合,将分类与定位任务分离,以提升模型收敛速度;此外引入CIoU作为瑕疵位置回归损失函数,提高瑕疵定位准确性.实验结果表明,本文算法在Raspberry Pi 4B上可达8.6 FPS的检测速度,可满足纺织工业应用需求.
