基于机器视觉、深度学习的涤纶丝锭外检  

External Inspection for Polyester Bobbins Based on Machine Vision and Deep Learning

作　　者：陈国文 周先何 叶勇 魏中青 倪国民 CHEN Guo-wen;ZHOU Xian-he;YE Yong;WEI Zhong-qing;NI Guo-min(Zhejiang Shengyuan Chemical Fiber Co.,Ltd.,Hangzhou 311247,Zhejiang,China)

机构地区：[1]浙江盛元化纤有限公司,浙江杭州1311247

出　　处：《合成纤维》2022年第12期41-44,共4页Synthetic Fiber in China

摘　　要：盛元涤纶牵伸变形丝锭智能外检系统利用光电成像系统采集被控目标的图像,经计算机及专用的图像处理模块进行数字化处理,根据图像的像素分布、亮度和颜色等信息,来进行尺寸、形状、色泽等的疵点判别。将深度学习的神经网络算法应用到涤纶丝锭外观缺陷的数据特征学习和识别上,克服了人眼无法连续工作、人为误判及对人员经验和熟练度依赖的缺点,稳定地完成这些带有高度重复性和缺陷收集难的工作。Photoelectric imaging system is used to collect the image of the controlled target in Shengyuan polyester fiber draw textured yarn(DTY) intelligent external inspection system, and then digital processing by computer and special image processing module is performed. According to the image pixel distribution,brightness and color information, the defect discrimination of size, shape and color is carried out. The neural network algorithm of deep learning is applied to the data feature learning and recognition of the appearance defects of polyester yarn bobbins, which overcomes the shortcomings of uncontinuous work of human eyes, artificial misjudgment and dependence on personnel experience and proficiency, and stably completes these tasks with high repeatability and difficult collection of defects.

关 键 词：涤纶丝锭 机器视觉 智能外检 深度学习 大数据 

分 类 号：TQ340.7[化学工程—化纤工业] TP39[自动化与计算机技术—计算机应用技术]