基于机器视觉的复杂环境下细纱落纱机的纱锭和管纱识别  

Recognition of spindles and tube yarns of spinning frames under complex environment based on machine vision

作　　者：闫小鹏 杨雅宁 YAN Xiaopeng;YANG Yaning(Scraper Research Branch,Ningxia Tiandi Benniu Industrial Group Co.,Ltd.,Shizuishan 753001,China;School of Physics and Electronic Information Engineering,Ningxia Normal University,Guyuan 756500,China)

机构地区：[1]宁夏天地奔牛实业集团有限公司刮板机研究分院,宁夏石嘴山753001 [2]宁夏师范学院物理与电子信息工程学院,宁夏固原756500

出　　处：《现代纺织技术》2024年第7期22-32,共11页Advanced Textile Technology

基　　金：固原市科技研发类计划项目(2021GYKYF005)。

摘　　要：为了解决传统自动细纱落纱机在拔管和插管操作中存在的漏拔、漏插问题,将机器视觉技术引入自动细纱落纱机中,提出了一种两次识别分割的方法,以准确识别管纱和纱锭,降低漏拔和漏插发生的概率。针对复杂环境对原始图像造成的严重干扰,首先在图像预处理时调整目标与背景之间的对比度,突出纱锭和管纱的图像特征;然后在目标识别过程中进行初始识别和二次识别。为进一步提高识别的准确率,分别在图像预处理和目标识别阶段进行改进,在预处理中通过图像乘法融合的方式进一步增强对比度,在目标筛选决策中引入目标数量特征以提高识别准确率。结果表明,该方法对不同场景下的纱锭和管纱的识别准确率超过98.40%,实现了在复杂背景下纱锭和管纱的精确识别。In the textile manufacturing industry,the process of yarn dropping in spinning frames is critical as it directly impacts yarn quality,production output,and factory operations.Traditional automatic doffers,while capable of performing operations automatically,often suffer from imprecise of operations,resulting in missed of empty tube insertion or tube yarn extraction during the doffing process.The current approach is to assign dedicated personnel to track,inspect,and address the issue,which increases labor costs and restricts the level of automation and intelligence in the workshop.This paper aimed to address the issues of yarn tube extraction missing and empty tube insertion missing in traditional automatic fine yarn doffers by introducing machine vision technology to accurately recognize tube yarns and yarn spindles,so as to reduce the probability of missed extractions and insertions in subsequent operations.However,traditional image processing techniques face challenges in complex fine yarn workshop environments,such as complex target backgrounds,lighting variations,obstructions,and shooting angles.In addition,it is necessary to set parameters and thresholds based on experience during the recognition process of spindles and tube yarns,which increases the difficulty of recognition.Therefore,accurate recognition of tube yarns and spindles is crucial for automatic doffers based on machine vision technology.To solve the interference problem of images gathered by visual sensors in workshop environments,this paper proposed a basic image processing method.By adjusting the contrast between the targets and the background,the characteristics of spindles and tube yarns could be highlighted.This method also performed initial and secondary recognition in the targets recognition process to obtain the target areas more accurately.In order to further improve the recognition accuracy of yarn spindles and tube yarns,the basic image processing method was improved in image preprocessing phase and object recognition phase,respective

关 键 词：机器视觉 落纱机 纱锭识别 管纱识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]