The exploration of smart electronic textiles is a common goal to improve people’s quality of life.However,current smart e-textiles still face challenges such as being prone to failure under humid or cold conditions,l...The exploration of smart electronic textiles is a common goal to improve people’s quality of life.However,current smart e-textiles still face challenges such as being prone to failure under humid or cold conditions,lack of washing durability and chemical fragility.Herein,a multifunctional strain sensor with a negative resistance change was developed based on the excellent elasticity of knitted fabrics.A reduced graphene oxide(rGO)conductive fabric was first obtained by electrostatic self-assembly of chitosan(CS).Then a strain sensor was prepared using a dip-coating process to adsorb nanoscale silica dioxide and poly(dimethylsiloxane)(PDMS).A broad working range of 60%,a fast response time(22 ms)and stable cycling durability over 4000 cycles were simultaneously achieved using the prepared sensor.Furthermore,the sensor showed excel-lent superhydrophobicity,photothermal effects and UV protection,as graphene,silica and PDMS acted in synergy.This multifunctional sensor could be mounted on human joints to perform tasks,including activity monitoring,medical rehabili-tation evaluation and gesture recognition,due to its superior electromechanical capabilities.Based on its multiple superior properties,this sensor could be used as winter sportswear for athletes to track their actions without being impacted by water and as a warmer to ensure the wearer's comfort.展开更多
为了满足工业上对织物缺陷检测的实时性要求,提出一种基于S-YOLOV3(Slimming You Only Look Once Version 3)模型的织物实时缺陷检测算法。首先使用K均值聚类算法确定目标先验框,以适应不同尺寸的缺陷;然后预训练YOLOV3模型得到权重参数...为了满足工业上对织物缺陷检测的实时性要求,提出一种基于S-YOLOV3(Slimming You Only Look Once Version 3)模型的织物实时缺陷检测算法。首先使用K均值聚类算法确定目标先验框,以适应不同尺寸的缺陷;然后预训练YOLOV3模型得到权重参数,利用批归一化层中的缩放因子γ评估每个卷积核的权重,将权重值低于阈值的卷积核进行剪枝以得到S-YOLOV3模型,实现模型压缩和加速;最后对剪枝后的网络进行微调以提高模型检测的准确率。实验结果表明:对于不同复杂纹理的织物,所提模型都能准确检测,且平均精度均值达到94%,剪枝后检测速度提高到55 FPS,所得的准确率与实时性均满足工业上的实际需求。展开更多
基金supported by the Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine(No.Z YYCXTD-D-202206)the Natural Science Foundation of Jiangxi Province,China(No.20212BAB214016)+3 种基金the Fundamental Research Funds for the Central Universities(No.JUSRP52007A)the International Science and Technology Center(No.BZ2018032)the Jiangsu Province Advanced Textile Engineering Technology Centre Funding Project(XJFZ/2021/4)the National Natural Science Foundation of China(No.51603090).
文摘The exploration of smart electronic textiles is a common goal to improve people’s quality of life.However,current smart e-textiles still face challenges such as being prone to failure under humid or cold conditions,lack of washing durability and chemical fragility.Herein,a multifunctional strain sensor with a negative resistance change was developed based on the excellent elasticity of knitted fabrics.A reduced graphene oxide(rGO)conductive fabric was first obtained by electrostatic self-assembly of chitosan(CS).Then a strain sensor was prepared using a dip-coating process to adsorb nanoscale silica dioxide and poly(dimethylsiloxane)(PDMS).A broad working range of 60%,a fast response time(22 ms)and stable cycling durability over 4000 cycles were simultaneously achieved using the prepared sensor.Furthermore,the sensor showed excel-lent superhydrophobicity,photothermal effects and UV protection,as graphene,silica and PDMS acted in synergy.This multifunctional sensor could be mounted on human joints to perform tasks,including activity monitoring,medical rehabili-tation evaluation and gesture recognition,due to its superior electromechanical capabilities.Based on its multiple superior properties,this sensor could be used as winter sportswear for athletes to track their actions without being impacted by water and as a warmer to ensure the wearer's comfort.
文摘为了满足工业上对织物缺陷检测的实时性要求,提出一种基于S-YOLOV3(Slimming You Only Look Once Version 3)模型的织物实时缺陷检测算法。首先使用K均值聚类算法确定目标先验框,以适应不同尺寸的缺陷;然后预训练YOLOV3模型得到权重参数,利用批归一化层中的缩放因子γ评估每个卷积核的权重,将权重值低于阈值的卷积核进行剪枝以得到S-YOLOV3模型,实现模型压缩和加速;最后对剪枝后的网络进行微调以提高模型检测的准确率。实验结果表明:对于不同复杂纹理的织物,所提模型都能准确检测,且平均精度均值达到94%,剪枝后检测速度提高到55 FPS,所得的准确率与实时性均满足工业上的实际需求。
