一种基于改进SSD的输液器导管涂胶缺陷检测方法
A Method for Detecting Adhesive Defects of Infusion Catheter Based on Improved SSD
谢骏杰 1陈瑞源 2王俊 1张石清 1罗坚 3郑龙3
作者信息
- 1. 台州学院电子与信息工程学院,浙江 台州 318000
- 2. 台州学院医学院,浙江 台州 318000
- 3. 迈得医疗工业设备股份有限公司,浙江 玉环 317607
- 折叠
摘要
在医疗器械的装配过程中,适量的导管涂胶可以有效避免导管的移位或脱落,为了确保输液治疗的安全性和稳定性,对导管涂胶的缺陷检测是医疗器械检测中必不可少的环节.然而现有的导管涂胶技术大多采用自动化控制胶水量的方法实现导管涂胶,缺少了涂胶缺陷检测的步骤,仍然存在导管涂胶不规范的隐患.一种基于深度学习的导管涂胶缺陷检测方法可以解决上述问题,该方法使用单发多框探测器(single shot multibox detector,SSD)目标检测算法实现对导管涂胶的缺陷检测.同时采用深度残差网络ResNet101 作为特征提取网络,提高了模型的特征建模能力,进一步提升了缺陷检测算法的准确性.实验结果表明,所提方法优于当前主流的缺陷检测算法,实现了高精度的导管涂胶缺陷检测,进一步推动了导管涂胶技术的发展.
Abstract
In the assembly process of medical devices,applying an appropriate amount of catheter glue can effective-ly avoid the displacement or detachment of the catheter.To ensure the safety and stability of infusion treatment,the de-fect detection of catheter glue is an indispensable part of medical device testing.However,most of the existing catheter gluing technologies use automated control of the amount of glue to realize the catheter gluing,lacking the steps of gluing defect detection,and there are still hidden dangers of non-standard catheter gluing.In order to solve the above prob-lems,this paper proposes a deep learning-based catheter coating defect detection,which uses SSD(Single Shot MultiBox Detector)object detection algorithm to achieve the defect detection of catheter coating.At the same time,ResNet101 is used as the feature extraction network,which improves the feature modeling ability of the model and further improves the accuracy of the defect detection algorithm.The experimental results show that the method proposed is superior to the current mainstream defect detection algorithms,achieving high precision defect detection of catheter coating,which further promotes the development of catheter coating technology.
关键词
导管涂胶/改进SSD/ResNet101/缺陷检测Key words
duct coating/improved SSD/ResNet101/defect detection引用本文复制引用
出版年
2024
NETL
NSTL
维普
万方数据