摘要
针对织物表面打印导电线路中传统的微滴目标检测通用性及智能化程度低的问题,使用深度学习模型实现了微滴形态的实时检测。通过对训练图片进行预处理,使用Labelimg工具标记不同形态的微滴灰度图像目标区域,生成相应的数据信息库,为后续的训练和测试提供数据集。利用深度学习目标检测模型训练标记好的数据集,得到最优检测模型。实验结果表明:相比于传统目标检测方法,深度学习检测模型对不同形态的微滴具有很好的适用性;相比于其他深度学习模型,Faster R-CNN模型的精度更高,对正常微滴和不良液滴检测的平均精度均值为84.89%。该研究为后续织物表面导电线路的精确成形提供了参考。
Due to the low versatility and intelligence of the traditional droplet target detection in the conductive circuit printed on the fabric surface, deep learning model was used to realize the real-time monitoring of droplet shape. By preprocessing the training images, the Labelimg tool was used to mark the target areas of the droplet grayscale images of different shapes, and the corresponding data information library was generated to provide data sets for subsequent training and testing. Using the labeled data set, the mainstream deep learning target detection models were trained separately to obtain the optimal detection model. The experimental results show that compared with traditional target detection methods, the deep learning detection model has good applicability to droplets of different shapes, and compared with other deep learning models, the Faster R-CNN model has higher accuracy, and the mean average precision(mAP) of normal and bad droplet detection is 84.89%. This research provides a reference for the subsequent precise formation of the circuit on the fabric surface.
作者
胥光申
徐晋
杨鹏程
肖渊
王兆辉
梁蒲佳
XU Guangshen;XU Jin;YANG Pengcheng;XIAO Yuan;WANG Zhaohui;LIANG Pujia(School of Mechanical and Electrical Engineering,Xi’an Polytechnic University,Xi’an 710048,China;Xi’an Key Laboratory of Modern Intelligent Textile Equipment,Xi’an 710600,China)
出处
《纺织高校基础科学学报》
CAS
2023年第1期41-48,56,共9页
Basic Sciences Journal of Textile Universities
基金
国家自然科学基金(51475350)
西安市现代智能纺织装备重点实验室建设项目(2019220614SYS021CG043)
西安工程大学研究生创新基金项目(chx2021008)。
关键词
智能纺织品
喷射打印
图像处理
目标检测
深度学习
smart textiles
jet printing
image processing
target detection
deep learning
